{"id":"7ded86c3-9435-4519-b841-2e761000effb","entityType":"agent","slug":"clawhub-parmasanandgarlic-farmdash-trail-marshal","name":"FarmDash Trail Marshal","canonicalUrl":"https://www.xpersona.co/agent/clawhub-parmasanandgarlic-farmdash-trail-marshal","canonicalPath":"/agent/clawhub-parmasanandgarlic-farmdash-trail-marshal","generatedAt":"2026-10-10T10:49:59.073Z","source":"CLAWHUB","claimStatus":"UNCLAIMED","verificationTier":"NONE","summary":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":null},"description":"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.","descriptionLabel":"Source description","evidenceSummary":"Capability contract not published. No trust telemetry is available yet. 1.6K downloads reported by the source. Last updated 10/10/2026.","installCommand":"clawhub skill install s177a1xxprmvcbyygraxwvbf5s83qzs6:farmdash-trail-marshal","sourceUrl":"https://clawhub.ai/parmasanandgarlic/farmdash-trail-marshal","homepage":"https://clawhub.ai/parmasanandgarlic/skills/farmdash-trail-marshal","primaryLinks":[{"label":"View on ClawHub","url":"https://clawhub.ai/parmasanandgarlic/farmdash-trail-marshal","kind":"source"},{"label":"Homepage","url":"https://clawhub.ai/parmasanandgarlic/skills/farmdash-trail-marshal","kind":"homepage"}],"safetyScore":84,"overallRank":62,"popularityScore":64,"trustScore":null,"claimedByName":null,"isOwner":false,"seoDescription":"FarmDash Trail Marshal technical dossier on Xpersona with agent coverage, OPENCLEW support, and live trust metadata."},"coverage":{"evidence":{"source":"public-profile","verified":false,"confidence":"medium","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":null},"protocols":[{"protocol":"OPENCLEW","label":"OpenClaw","status":"self-declared","notes":"Declared in the public agent profile."}],"capabilities":[],"verifiedCount":0,"selfDeclaredCount":1,"capabilityMatrix":{"rows":[{"key":"OPENCLEW","type":"protocol","support":"unknown","confidenceSource":"profile","notes":"Listed on profile"}],"flattenedTokens":"protocol:OPENCLEW|unknown|profile"}},"adoption":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":null},"stars":null,"forks":null,"downloads":1628,"packageName":null,"latestVersion":"0.1.10","tractionLabel":"1.6K downloads"},"release":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":null},"lastUpdatedAt":"2026-10-10T06:10:56.163Z","lastCrawledAt":"2026-10-10T06:10:56.163Z","lastIndexedAt":null,"nextCrawlAt":"2026-10-11T06:10:56.163Z","lastVerifiedAt":null,"highlights":[{"version":"0.1.10","createdAt":"2026-09-26T23:45:29.903Z","changelog":"- Removed the documentation file skill-card.md. - No functional or interface changes to the skill itself. - All workflow tools and catalog features remain unchanged.","fileCount":4,"zipByteSize":11171},{"version":"0.1.9","createdAt":"2026-09-07T18:05:53.753Z","changelog":"- Removed the sample skill-card.md file for cleaner packaging. - Updated SKILL.md description and tags for clarity—emphasized policy-gated, supervised workflows and concise DeFi use hints. - Expanded catalog guidance in SKILL.md to make recipe selection for common DeFi goals easier. - Version bump to 1.1.2.","fileCount":4,"zipByteSize":11227},{"version":"0.1.8","createdAt":"2026-08-04T05:02:21.262Z","changelog":"FarmDash Trail Marshal 0.1.8 Changelog - Removed the file: skill-card.md - No functional or feature changes; documentation-only cleanup.","fileCount":4,"zipByteSize":10114},{"version":"0.1.7","createdAt":"2026-06-28T01:34:53.625Z","changelog":"- Removed the file: skill-card.md - No feature or user-facing changes; this is a maintenance update. - All functionality and documentation remain unchanged.","fileCount":4,"zipByteSize":9262},{"version":"0.1.6","createdAt":"2026-05-30T23:52:23.604Z","changelog":"farmdash-trail-marshal 0.1.6 - Updated metadata to add `FARMDASH_API_KEY` environment variable and clarify API key requirements. - Improved documentation to specify API token options, including public scout access without a key. - The skill-card.md file was removed; SKILL.json was added for updated metadata. - Version and metadata fields updated for clarity and alignment.","fileCount":4,"zipByteSize":9446},{"version":"0.1.5","createdAt":"2026-05-24T22:38:11.401Z","changelog":"No user-facing changes detected in this version. - No file changes were made. - Functionality and documentation remain unchanged from the previous release.","fileCount":3,"zipByteSize":7888},{"version":"0.1.4","createdAt":"2026-05-22T17:17:01.569Z","changelog":"- Expanded workflow catalog from 14 to 18 recipes, adding: - `pre_trade_edge_audit` - `post_execution_quality_review` - `perps_liquidation_buffer_check` - `funding_carry_break_even_audit` - Updated tags for broader search and ecosystem alignment. - Documentation clarifications; no functional or code changes detected.","fileCount":2,"zipByteSize":6309},{"version":"0.1.3","createdAt":"2026-05-20T21:45:53.013Z","changelog":"**Major update: FarmDash Trail Marshal moves from static cookbook to guarded orchestration and workflow session management.** - Adds new tools: `plan_workflow`, `run_workflow`, and `get_workflow_status` for session-based guarded orchestration. - Workflow quality gates classify available, missing, and state-changing steps based on installed skills before execution. - Workflow run records can now be created and tracked per session; workflows requiring execution now move to `awaiting_confirmation` until user approval via the appropriate sub-skill. - Expanded catalog to 14 orchestrated workflow recipes with updated step metadata. - Trail Marshal still performs no on-chain execution and only interacts with user-approved, separately-installed execution skills.","fileCount":2,"zipByteSize":5994}]},"execution":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No published capability contract is available yet."},"installCommand":"clawhub skill install s177a1xxprmvcbyygraxwvbf5s83qzs6:farmdash-trail-marshal","setupComplexity":"low","setupSteps":["Install using `clawhub skill install s177a1xxprmvcbyygraxwvbf5s83qzs6:farmdash-trail-marshal` in an isolated environment before connecting it to live workloads.","No published capability contract is available yet, so validate auth and request/response behavior manually.","Review the upstream CLAWHUB listing at https://clawhub.ai/parmasanandgarlic/farmdash-trail-marshal before using production credentials."],"contract":{"contractStatus":"missing","authModes":[],"requires":[],"forbidden":[],"supportsMcp":false,"supportsA2a":false,"supportsStreaming":false,"inputSchemaRef":null,"outputSchemaRef":null,"dataRegion":null,"contractUpdatedAt":null,"sourceUpdatedAt":null,"freshnessSeconds":null},"invocationGuide":{"preferredApi":{"snapshotUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/trust"},"curlExamples":["curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/snapshot\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/contract\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/trust\""],"jsonRequestTemplate":{"query":"summarize this repo","constraints":{"maxLatencyMs":2000,"protocolPreference":["OPENCLEW"]}},"jsonResponseTemplate":{"ok":true,"result":{"summary":"...","confidence":0.9},"meta":{"source":"CLAWHUB","generatedAt":"2026-10-10T10:49:59.069Z"}},"retryPolicy":{"maxAttempts":3,"backoffMs":[500,1500,3500],"retryableConditions":["HTTP_429","HTTP_503","NETWORK_TIMEOUT"]}},"endpoints":{"dossierUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/dossier","snapshotUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-parmasanandgarlic-farmdash-trail-marshal/trust"}},"reliability":{"evidence":{"source":"runtime-metrics","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No trust, reliability, or runtime telemetry is available."},"trust":{"status":"unavailable","handshakeStatus":"UNKNOWN","verificationFreshnessHours":null,"reputationScore":null,"p95LatencyMs":null,"successRate30d":null,"fallbackRate":null,"attempts30d":null,"trustUpdatedAt":null,"trustConfidence":"unknown","sourceUpdatedAt":null,"freshnessSeconds":null},"decisionGuardrails":{"doNotUseIf":["Contract metadata is missing or unavailable for deterministic execution."],"safeUseWhen":[],"riskFlags":["missing_or_unavailable_contract","trust_data_unavailable","schema_references_missing"],"operationalConfidence":"low"},"executionMetrics":{"observedLatencyMsP50":null,"observedLatencyMsP95":null,"estimatedCostUsd":null,"uptime30d":null,"rateLimitRpm":null,"rateLimitBurst":null,"lastVerifiedAt":null,"verificationSource":null},"runtimeMetrics":{"successRate":null,"avgLatencyMs":null,"avgCostUsd":null,"hallucinationRate":null,"retryRate":null,"disputeRate":null,"p50Latency":null,"p95Latency":null,"lastUpdated":null}},"benchmarks":{"evidence":{"source":"no-benchmark-data","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No benchmark suites or observed failure patterns are available."},"suites":[],"failurePatterns":[]},"artifacts":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":null},"readme":"Skill: FarmDash Trail Marshal\n\nOwner: parmasanandgarlic\n\nSummary: Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\n\nTags: agent:0.1.1, agent-orchestration:0.1.9, agent-recipes:0.1.6, agent-workflows:0.1.7, agentic defi:0.1.10, ai agent:0.1.10, ai agent orchestration layer:0.1.10, ai agent risk management:0.1.10, ai-agent:0.1.7, ai-agent-workflows:0.1.9, airdrop:0.1.7, airdrop farming rotation:0.1.10, airdrop-workflow:0.1.9, anthropic:0.1.1, automation:0.1.7, autonomous-agent:0.1.7, claude:0.1.1, claude defi skill:0.1.10, clawhub:0.1.7, composer:0.1.0, confirmation gated defi:0.1.10, crypto:0.1.7, crypto tax loss harvesting:0.1.10, defi:0.1.10, defi automation:0.1.10, defi emergency exit plan:0.1.10, defi workflow quality gate:0.1.10, defi-automation:0.1.8, defi-playbooks:0.1.9, defi-workflows:0.1.9, delta neutral defi strategy:0.1.10, execution-gate:0.1.6, farmdash:0.1.9, funding rate arbitrage tool:0.1.10, guarded-orchestration:0.1.5, hyperliquid:0.1.5, hyperliquid ai agent:0.1.10, hyperliquid points farming:0.1.10, langgraph:0.1.10, langgraph defi agent:0.1.10, latest:0.1.10, mcp:0.1.9, mcp defi workflows:0.1.10, mcp server:0.1.10, model context protocol:0.1.10, multi-agent:0.1.7, multi-agent-system:0.1.9, multi-skill:0.1.6, no-custody:0.1.1, non-custodial defi:0.1.10, onchain:0.1.7, openclaw:0.1.9, openclaw skill:0.1.10, openclaw skill store:0.1.10, orchestration:0.1.7, perps liquidation buffer check:0.1.10, points-farming:0.1.5, policy gated ai agent:0.1.10, policy-gated:0.1.9, portfolio:0.1.5, portfolio rebalancing ai:0.1.10, post trade reconciliation crypto:0.1.10, pre trade edge audit:0.1.10, quality-gates:0.1.9, quarterly portfolio rotation defi:0.1.10, read only defi research agent:0.1.10, read-only:0.1.7, research:0.1.1, risk-management:0.1.9, session-management:0.1.9, stablecoin yield ladder:0.1.10, state-machine:0.1.7, supabase ai agent:0.1.10, supervised ai automation:0.1.10, supervised-agent:0.1.5, supervised-automation:0.1.9, task-planning:0.1.6, trail-marshal:0.1.7, web3:0.1.9, web3 automation:0.1.10, workflow:0.1.7, workflow-automation:0.1.9, workflow-planning:0.1.5, workflow-status:0.1.9, yield-farming:0.1.7, yield-farming-workflow:0.1.9, zero custody ai agent:0.1.10, zero-custody:0.1.9\n\nVersion history:\n\nv0.1.10 | 2026-09-26T23:45:29.903Z | user\n\n- Removed the documentation file skill-card.md.\n- No functional or interface changes to the skill itself.\n- All workflow tools and catalog features remain unchanged.\n\nv0.1.9 | 2026-09-07T18:05:53.753Z | user\n\n- Removed the sample skill-card.md file for cleaner packaging.\n- Updated SKILL.md description and tags for clarity—emphasized policy-gated, supervised workflows and concise DeFi use hints.\n- Expanded catalog guidance in SKILL.md to make recipe selection for common DeFi goals easier.\n- Version bump to 1.1.2.\n\nv0.1.8 | 2026-08-04T05:02:21.262Z | user\n\nFarmDash Trail Marshal 0.1.8 Changelog\n\n- Removed the file: skill-card.md\n- No functional or feature changes; documentation-only cleanup.\n\nv0.1.7 | 2026-06-28T01:34:53.625Z | user\n\n- Removed the file: skill-card.md\n- No feature or user-facing changes; this is a maintenance update.\n- All functionality and documentation remain unchanged.\n\nv0.1.6 | 2026-05-30T23:52:23.604Z | user\n\nfarmdash-trail-marshal 0.1.6\n\n- Updated metadata to add `FARMDASH_API_KEY` environment variable and clarify API key requirements.\n- Improved documentation to specify API token options, including public scout access without a key.\n- The skill-card.md file was removed; SKILL.json was added for updated metadata.\n- Version and metadata fields updated for clarity and alignment.\n\nv0.1.5 | 2026-05-24T22:38:11.401Z | user\n\nNo user-facing changes detected in this version.\n\n- No file changes were made.\n- Functionality and documentation remain unchanged from the previous release.\n\nv0.1.4 | 2026-05-22T17:17:01.569Z | user\n\n- Expanded workflow catalog from 14 to 18 recipes, adding:\n  - `pre_trade_edge_audit`\n  - `post_execution_quality_review`\n  - `perps_liquidation_buffer_check`\n  - `funding_carry_break_even_audit`\n- Updated tags for broader search and ecosystem alignment.\n- Documentation clarifications; no functional or code changes detected.\n\nv0.1.3 | 2026-05-20T21:45:53.013Z | user\n\n**Major update: FarmDash Trail Marshal moves from static cookbook to guarded orchestration and workflow session management.**\n\n- Adds new tools: `plan_workflow`, `run_workflow`, and `get_workflow_status` for session-based guarded orchestration.\n- Workflow quality gates classify available, missing, and state-changing steps based on installed skills before execution.\n- Workflow run records can now be created and tracked per session; workflows requiring execution now move to `awaiting_confirmation` until user approval via the appropriate sub-skill.\n- Expanded catalog to 14 orchestrated workflow recipes with updated step metadata.\n- Trail Marshal still performs no on-chain execution and only interacts with user-approved, separately-installed execution skills.\n\nv0.1.2 | 2026-05-20T03:49:45.962Z | user\n\nFarmDash Trail Marshal v0.2.0 – Catalog and API update\n\n- Updated the workflow catalog: revised recipe set, improved descriptions, and step details.\n- Unified workflow filtering: legacy `category`/`tier` filters are now a general `filter` string.\n- Updated step syntax: each workflow lists steps as `{ skill, tool, purpose }` entries; obsolete aliases (e.g., `get_balances`) are removed.\n- Introduced workflow Quality Gate pattern: classify missing/execution steps, safe subsets, and enforce fallback for incomplete tooling.\n- Tier pricing and access updated: Pioneer tier is now $29/mo and filter support follows API changes.\n- Improved agent guidance on catalog caching and skill availability handling.\n\nv0.1.1 | 2026-05-13T15:51:38.470Z | user\n\nFarmDash Trail Marshal v0.1.1\n\n- Converts to a strictly read-only metadata cookbook (“static catalog” model).\n- Removes all environment variables and credential surfaces; no keys or tokens required.\n- Now exports one tool only: list_workflows — returns a canonical set of named workflow recipes, each detailing its required user confirmations and sub-skill handoffs.\n- All workflow execution, signature, and state change is explicitly delegated to the user’s separately-installed sub-skills under their own contracts.\n- Updates documentation for agent guidance, security stance, and risk warnings.\n- Reorganizes and clarifies tier model, recipe summaries, and permission boundaries.\n\nv0.1.0 | 2026-05-13T15:38:31.059Z | user\n\nFarmDash Trail Marshal v0.1.0 — initial release\n\n- Introduces Trail Marshal as a read-only workflow composition/catalog layer for FarmDash agents.\n- Publishes a canonical workflow catalog via the `list_workflows` tool, enabling deterministic multi-skill orchestration.\n- Supports cross-skill workflows spanning Trail Intelligence, Wagon Steward, Signal Architect, and Futures Strategist.\n- No transaction/signing; does not analyze or execute, only guides agents on workflow sequencing and confirmation.\n- Ideal when user goals require coordination of more than one FarmDash skill or an end-to-end farming plan.\n\nArchive index:\n\nArchive v0.1.10: 4 files, 11171 bytes\n\nFiles: skill-card.md (2266b), SKILL.json (5231b), SKILL.md (20483b), _meta.json (142b)\n\nFile v0.1.10:SKILL.md\n\n---\nname: FarmDash Trail Marshal\ndescription: \"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\"\ntags: [\"defi\",\"defi-workflows\",\"agent-orchestration\",\"workflow-automation\",\"multi-agent-system\",\"ai-agent-workflows\",\"policy-gated\",\"supervised-automation\",\"defi-playbooks\",\"quality-gates\",\"workflow-status\",\"session-management\",\"risk-management\",\"yield-farming-workflow\",\"airdrop-workflow\",\"zero-custody\",\"openclaw\",\"mcp\",\"web3\",\"farmdash\"]\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\nhomepage: https://www.farmdash.one/agents\nversion: \"1.1.2\"\nicon: 🪪\nenv:\n  FARMDASH_API_KEY:\n    description: \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\"\n    required: false\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"primaryEnv\":\"FARMDASH_API_KEY\",\"apiKeyRequired\":false,\"freeScoutKeyless\":true,\"freeScoutKey\":\"fd_scout_free\",\"execution\":\"guarded-orchestration-no-execution\"}}\n---\n\n# FarmDash Trail Marshal\n\n> Use this skill when running a multi-step DeFi goal: pick a named, confirmation-gated recipe instead of improvising the sequence at runtime.\n\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\n\n## What this skill does\n\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\n\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\n\n## Available tools\n\n### `list_workflows`\n\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\n\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\n\n**Returns shape:**\n\n```\n{\n  \"workflows\": [\n    {\n      \"id\": \"farm_hyperliquid\",\n      \"name\": \"Hyperliquid Point Farming\",\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\n      \"requiredTier\": \"syndicate\",\n      \"steps\": [\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\n      ]\n    }\n  ]\n}\n```\n\nThe detailed `steps` for each recipe are delivered at runtime as part of the `list_workflows` JSON response. Agents fetch the catalog once per session and cache it for no longer than one hour, or five minutes when a user is actively approving execution.\n\nCodebase alignment note: Wagon steps use `get_wallet_balances`; do not use the stale alias `get_balances`.\n\n**Agent posture:** Call `list_workflows` once at conversation start. Present recipes by name when the user describes a goal that matches one of them.\n\n### `plan_workflow`\n\nBuilds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\n\n- available steps.\n- missing steps.\n- state-changing steps.\n- confirmation count.\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\n\nUse this before telling the user a workflow is executable.\n\n### `run_workflow`\n\nCreates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\n\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\n\n### `get_workflow_status`\n\nReads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\n\n## Workflow catalog (high level)\n\n### Which Recipe for Which Goal\nIdle stables or ladder intent: `stablecoin_yield_ladder`, then `idle_capital_deploy`. Existing-book review: `rebalance_portfolio`, `rotate_quarterly`, or `yield_optimization`. Risk or exit: `protect_portfolio` for scans, `emergency_exit` for unwinds. Post-trade: `post_trade_ledger_review` then `post_execution_quality_review`. Pre-trade validation: `pre_trade_edge_audit`, `funding_carry_break_even_audit`, `perps_liquidation_buffer_check`. Continuous sessions: `session_command_center`. Never use `sybil_dilution` for transactions.\n\nEighteen recipes are published. Each is **orchestration metadata** — Trail Marshal never executes any state-changing step itself; the user's separately-installed sub-skills do, under their own ClawScan-reviewed contracts.\n\n| ID | Goal | Tier | User confirmations |\n|---|---|---|---|\n| `farm_hyperliquid` | Coordinate Hyperliquid points, spot exposure, and perps/funding research | Syndicate | per execution step |\n| `rotate_quarterly` | Compare current positions against fresh high-heat opportunities | Pioneer | per swap |\n| `protect_portfolio` | Risk-first scan and optional emergency mitigation | Pioneer | 0 if no action |\n| `tax_loss_harvest` | Identify loss-harvest candidates and quote realization paths | Pioneer | per swap |\n| `delta_neutral_setup` | Plan paired spot + perp exposure with non-atomic legging controls | Syndicate | research-only until a paired execution adapter exists |\n| `idle_capital_deploy` | Find idle stables/native assets and compare deployment targets | Scout catalog / Pioneer wallet data | per swap |\n| `sybil_dilution` | Compatibility ID for read-only Automation Compliance Review; no evasion transactions | Pioneer | 0 |\n| `yield_optimization` | Review harvest/redeposit candidates against Trail Heat | Pioneer | per swap |\n| `emergency_exit` | Plan a protocol exit and unwind route | Scout catalog / Pioneer wallet data | per swap |\n| `rebalance_portfolio` | Re-align portfolio to target weights and current Trail Heat | Pioneer | per swap |\n| `stablecoin_yield_ladder` | Build a risk-tiered stablecoin deployment ladder | Pioneer | per deposit intent |\n| `post_trade_ledger_review` | Reconcile activity and export records after execution | Pioneer | 0 |\n| `pre_trade_edge_audit` | Confirm positive net edge, clean route quality, fresh risk checks, and quote stability before execution | Pioneer | 0 before execution handoff |\n| `post_execution_quality_review` | Compare expected versus realized output before allowing dependent follow-on actions | Pioneer | 0 |\n| `perps_liquidation_buffer_check` | Reject or resize perps orders that do not survive normal volatility around liquidation | Pioneer | 0 before execution handoff |\n| `funding_carry_break_even_audit` | Validate funding trades after fees, slippage, margin pressure, and funding flip risk | Pioneer | 0 before execution handoff |\n| `session_command_center` | Coordinate context, event snapshots, workflow plans, and heartbeats | Pioneer | per execution handoff |\n| `hedged_airdrop_rotation` | Rotate into high-heat farms while planning hedge coverage | Syndicate | spot leg + hedge leg |\n\nEach recipe's full step graph is returned at runtime by `list_workflows`. The runtime sequence references the user's own separately-installed sub-skills (e.g. *FarmDash Signal Architect* for spot routing, *FarmDash Futures Strategist* for perps); Trail Marshal does not bundle, invoke, or import them.\n\n`sybil_dilution` is a legacy identifier only. Never generate transactions to imitate organic activity, dilute clustering signals, or evade protocol anti-abuse controls.\n\n## Permissions\n\nTrail Marshal can read the public workflow catalog without session state. `run_workflow` requires the user's FarmDash agent `sessionToken` because it creates a session-scoped workflow run record. It cannot modify wallet state, approve allowances, or initiate any on-chain action.\n\nIf a recipe in the catalog requires execution, the user's separately-installed sub-skill performs that work under its own ClawScan-reviewed contract — Trail Marshal stays out of the execution call path.\n\nIf the runtime catalog references a skill that is not installed in the user's environment, such as `farmdash-camp-guard`, `farmdash-supply-master`, `farmdash-hedge-warden`, or `farmdash-ledger-keeper`, mark that step as unavailable and continue with the safe subset. Never fabricate a missing tool call.\n\n## Tier model\n\n| Tier | Cost | Limits | Capability |\n|---|---|---|---|\n| **Scout** | Free | 5 req / 24h | Returns the public workflow catalog |\n| **Pioneer** | $39.99/mo | 1,500 req / day | Adds higher-rate catalog use; filtering is via the `filter` search string |\n| **Syndicate** | $199/mo | 50k req / day | Teams, serious agents, high-volume orchestration, webhooks, unrestricted CORS, advanced session/control tooling |\n\n## Composition patterns\n\n### Workflow Quality Gate (v0.2)\n\n### Gate Before You Present\nRun `plan_workflow` with `workflowId`, `installedSkills`, and `agentAddress` before calling any recipe executable. Executable requires every state-changing step's owning skill installed; a missing execution step forces `analysis_only`, a missing read-only step lowers confidence. Quote the confirmation count up front as per execution step unless the live catalog is stricter.\n\nBefore presenting a recipe as executable, classify every step:\n\n```json\n{\n  \"workflowId\": \"rotate_quarterly\",\n  \"availableSteps\": [],\n  \"missingSteps\": [],\n  \"stateChangingSteps\": [],\n  \"confirmationsRequired\": 0,\n  \"safeSubset\": true,\n  \"fallback\": \"research_only | analysis_only | halt\"\n}\n```\n\nRules:\n\n- A workflow is executable only when every state-changing step has its owning execution skill installed.\n- A missing read-only step lowers confidence; a missing execution step changes the recipe to `analysis_only`.\n- If the recipe includes `execute_swap` or `execute_perp_order`, quote the confirmation count as \"per execution step\" unless the live catalog supplies a stricter count.\n- If more than five minutes pass after the user reviews a plan, re-fetch `list_workflows` and re-run the read-only sense steps.\n- A catalog step is not evidence that the named tool supplies every field in its purpose. Validate live tool contracts and downgrade claims when a compatibility alias is narrower than its name.\n- Never treat submitted/broadcast as filled/confirmed, or a scalar score as a probability.\n- Multi-leg execution is non-atomic unless the adapter proves otherwise. Define leg order, maximum unhedged time, partial-fill handling, abort conditions, and unwind before the first signature.\n\n### The Sense / Decide / Present / Verify loop\n\nEvery Trail Marshal recipe respects this 4-phase pattern:\n\n```\nSENSE    → Read-only research and portfolio skills gather state\nDECIDE   → The recipe ranks options and presents a plan to the user\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle\n```\n\nIf a recipe skips SENSE, it is unsafe. If it skips VERIFY, the agent never improves. Trail Marshal recipes always include both.\n\n### Confirmation gate\n\n### Multi-Leg Execution Disclosure\nTell the user multi-leg execution is non-atomic unless the adapter proves otherwise, and state leg order, maximum unhedged time, partial-fill handling, abort conditions, and unwind before the first signature. Trail Marshal cannot pre-confirm anything; each confirmation happens in the owning sub-skill's quote path, and `delta_neutral_setup` stays research-only until a paired adapter exists.\n\nTrail Marshal cannot pre-confirm anything. Each user confirmation in a recipe happens at the moment the user's *separate* sub-skill presents its quote — not in Trail Marshal's call path. If asked to run a recipe without intermediate review, the agent should refuse and explain the contract.\n\nFor `delta_neutral_setup`, the current catalog intentionally stops before execution. Neither `execute_swap` nor `execute_perp_order` can atomically bind both legs, and standalone `funding_arb` execution is analysis-only. A future paired adapter must bind both legs, maximum basis/slippage, timeout, partial-fill policy, and unwind before this workflow becomes executable.\n\n### Quant workflow gate\n\nBefore presenting any recipe as action-ready, require objective/horizon, source timestamps and freshness, full cost stack, conservative net edge, downside/invalidation, portfolio concentration impact, and missing evidence. Unknown is not zero. Trail Heat, FarmScore, yield sustainability, sybil risk, and futures confidence use different methodologies and must never be multiplied into a synthetic probability.\n\nFail to `research_only` or `halt` when data is stale/degraded, a critical safety flag is unresolved, authoritative settlement is missing for a dependency, or an execution step cannot bind the parameters the user reviewed.\n\n### Tier composition\n\nA recipe's required tier is the **maximum** of its sub-skill tiers. When a user is one tier below a recipe's requirement, the agent should recognize the gap from `requiredTier`, offer the lower-tier subset of the recipe (e.g. research-only without the execution leg), and surface an upgrade link if the user wants the full version. Never silently downgrade a recipe without telling the user what is being skipped.\n\n## Reasoning guidelines for agents\n\n- **Speak in workflows, not tool soup.** Quote a recipe's name and goal when presenting a plan.\n- **Quote the `confirms` count up front** so the user knows how many user-confirmation steps the recipe involves.\n- **Re-fetch and re-present** if more than ~5 minutes pass between catalog read and the user's review.\n\n### Catalog Freshness Discipline\nCache `list_workflows` for no longer than one hour, or five minutes once the user is actively approving. If more than five minutes pass after the user reviews a plan, re-fetch the catalog, re-run the read-only Sense steps, and re-present before any handoff to an execution sub-skill.\n- **No new analysis.** Trail Marshal returns recipe metadata; it does not editorialize or override what other skills produce.\n\n## Risk warnings the agent should surface\n\nFor any recipe whose execution legs are owned by a separately-installed sub-skill, the agent should restate to the user that:\n\n- DeFi positions can lose value rapidly; airdrop rewards are speculative and not guaranteed.\n- Smart contracts can be exploited; even high-Trail-Heat protocols carry technical risk.\n- Cross-chain transfers are irreversible if the wrong address or chain is selected.\n- Slippage, gas, MEV, and routing fees materially affect realized returns.\n- Past Trail Heat performance does not guarantee future scores.\n- The user retains full responsibility for every on-chain decision they confirm via their own execution skill.\n\n## Commercial disclosure (inherited from sub-skills)\n\nWhen a recipe surfaces a `https://www.farmdash.one/go/{slug}` partnership route, the standard FarmDash commercial disclosure must be included in the same message: FarmDash may receive referral, affiliate, or routing compensation, and fee details live at [FarmDash Fee Structure](https://www.farmdash.one/fees). Trail Marshal does not introduce new partner routes of its own.\n\n## Versioning\n\n**v1.0 (current).** Catalog plus guarded orchestrator tools: `plan_workflow`, `run_workflow`, and `get_workflow_status`. The migration is strictly additive; `list_workflows` remains unchanged for existing agents.\n\n## Disclaimers\n\nWorkflow recipes are documentation, not financial advice. Any action performed by separately-installed sub-skills is the user's decision under that sub-skill's own contract. Trail Marshal performs no on-chain action; the user retains full responsibility for every decision they confirm.\n\n---\n\n**Install:** Copy this file into your OpenClaw workspace, or fetch `https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md`.\n\n**Companion skills (separately installed by the user):** FarmDash Trail Intelligence (research), FarmDash Wagon Steward (read-only portfolio), FarmDash Signal Architect (spot routing), FarmDash Futures Strategist (perps), FarmDash Camp Guard (risk), FarmDash Supply Master (yield), FarmDash Hedge Warden (hedges), FarmDash Ledger Keeper (records), FarmDash Autonomous Operator (session context).\n\n**Why FarmDash:** Unlike single-prompt agent chains, Trail Marshal runs named recipes with per-step confirmation counts, separates read-only steps from state-changing ones, and refuses to invent sequences at runtime.\n\n**FarmDash:** [Autonomous DeFi workflow intelligence](https://www.farmdash.one/)\n**Agent Hub:** [FarmDash autonomous DeFi workflow platform](https://www.farmdash.one/agents)\n**OpenAPI Spec:** [FarmDash API Schema](https://www.farmdash.one/agents/openapi.yaml)\n**MCP Config:** [FarmDash MCP Server](https://www.farmdash.one/.well-known/mcp.json)\n\n---\n\n## Optional Setup Check\n\nNo registration call is required to install this skill, browse workflow recipes, or use research-only planning. Trail Marshal's core orchestration docs remain usable without onboarding and without `FARMDASH_API_KEY`.\n\nOnly run the setup check if the user explicitly asks to verify FarmDash tier/setup status and agrees to send the listed metadata.\n\nData sent:\n\n- public `agentAddress`\n- `X-ClawHub-Skill: farmdash-trail-marshal`\n\nData not sent:\n\n- private keys, seed phrases, mnemonics, wallet exports, OAuth tokens, raw wallet secrets, or signed transactions\n\nOptional command after consent:\n\n```bash\ncurl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\n  -H \"Content-Type: application/json\" \\\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'\n```\n\nThis returns tier status and available tool access. Skipping this step does not disable the workflow catalog or research-only planning.\n\n### Next steps:\n1. Use the Scout workflow catalog without setup when no key is configured.\n2. Add `FARMDASH_API_KEY` only when the user wants higher-rate Pioneer/Syndicate workflow features.\n3. Browse the OpenAPI spec at `https://www.farmdash.one/agents/openapi.yaml`.\nSend `X-ClawHub-Skill: farmdash-trail-marshal` on every FarmDash API request — not just the optional check above — so anonymous Scout usage is attributed to this skill. Analytics-only and optional; requests without the header still work. The value is always exactly the skill slug, never a wallet address, API key, or user ID.\n\n<!-- farmdash-canonical-links:start -->\n\n## Official FarmDash Links\n\n- [FarmDash DeFi intelligence website](https://www.farmdash.one/)\n- [FarmDash Agent Hub](https://www.farmdash.one/agents)\n- [Canonical FarmDash Trail Marshal skill manual](https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md)\n- [Agent integration documentation](https://www.farmdash.one/docs)\n- [Live agent capability status](https://www.farmdash.one/api/v1/agent/status)\n- [OpenAPI contract](https://www.farmdash.one/agents/openapi.yaml)\n- [MCP discovery manifest](https://www.farmdash.one/.well-known/mcp.json)\n- [Fees and commercial terms](https://www.farmdash.one/fees)\n- [Security and authority boundaries](https://www.farmdash.one/security)\n\n<!-- farmdash-canonical-links:end -->\n\nFile v0.1.10:_meta.json\n\n{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.10\",\n  \"publishedAt\": 1790466329903\n}\n\nFile v0.1.10:skill-card.md\n\n## Description:\n\nCoordinates named DeFi workflow recipes, checks required skills, and records guarded runs without executing wallet transactions.\n\nThis skill is ready for commercial/non-commercial use.\n\n## Publisher:\n\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic)\n\n### License/Terms of Use:\n\nMIT-0\n\n## Use Case:\n\nDeFi users and their agents use this skill to choose named workflow recipes, assess missing prerequisites and confirmation requirements, and track supervised runs. Wallet-changing steps require separately installed execution skills and explicit user confirmation.\n\n### Deployment Geography for Use:\n\nGlobal\n\n## Known Risks and Mitigations:\n\nRisk: A workflow may suggest financially risky DeFi actions or rely on stale market data.\n\nMitigation: Refresh research and review costs and risks before any handoff; require separate user confirmation for every wallet-changing action.\n\nRisk: Sharing wallet secrets or credentials with an untrusted endpoint could compromise assets or account access.\n\nMitigation: Never provide private keys, seed phrases, signed transactions, or raw wallet secrets; use optional API keys and session tokens only with trusted FarmDash endpoints.\n\n## Reference(s):\n\n- [FarmDash Agent Hub](https://www.farmdash.one/agents)\n- [FarmDash agent integration documentation](https://www.farmdash.one/docs)\n- [FarmDash OpenAPI contract](https://www.farmdash.one/agents/openapi.yaml)\n- [FarmDash MCP discovery manifest](https://www.farmdash.one/.well-known/mcp.json)\n- [FarmDash Trail Marshal skill manual](https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md)\n\n## Skill Output:\n\n**Output Type(s):** [Guidance, Workflow plans, Workflow status]\n\n**Output Format:** [Text and structured JSON]\n\n**Output Parameters:** [1D]\n\n**Other Properties Related to Output:** [Catalog entries, prerequisite and confirmation checks, and guarded run records; no wallet execution.]\n\n## Skill Version(s):\n\n0.1.10 (source: server-resolved ClawHub release)\n\n## Ethical Considerations:\n\nUsers should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment.\n\nFile v0.1.10:SKILL.json\n\n{\n  \"name\": \"FarmDash Trail Marshal\",\n  \"version\": \"1.1.2\",\n  \"description\": \"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\",\n  \"tags\": [\n    \"defi\",\n    \"defi-workflows\",\n    \"agent-orchestration\",\n    \"workflow-automation\",\n    \"multi-agent-system\",\n    \"ai-agent-workflows\",\n    \"policy-gated\",\n    \"supervised-automation\",\n    \"defi-playbooks\",\n    \"quality-gates\",\n    \"workflow-status\",\n    \"session-management\",\n    \"risk-management\",\n    \"yield-farming-workflow\",\n    \"airdrop-workflow\",\n    \"zero-custody\",\n    \"openclaw\",\n    \"mcp\",\n    \"web3\",\n    \"farmdash\"\n  ],\n  \"author\": \"FarmDash Pioneers (@Parmasanandgarlic)\",\n  \"homepage\": \"https://www.farmdash.one/agents\",\n  \"icon\": \"🪪\",\n  \"tools\": [\n    {\n      \"name\": \"list_workflows\",\n      \"description\": \"Returns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\\r\\n\\r\\n**\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"filter\": {\n            \"type\": \"string\"\n          }\n        },\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"plan_workflow\",\n      \"description\": \"Builds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\\r\\n\\r\\n- available steps.\\r\\n- missing steps.\\r\\n- state-changing steps.\\r\\n- confirmation count.\\r\\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\\r\\n\\r\\nUse this before telling the user a workflow is executable.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"workflowId\": {\n            \"type\": \"string\"\n          },\n          \"installedSkills\": {\n            \"type\": \"array\",\n            \"items\": {\n              \"type\": \"string\"\n            }\n          },\n          \"agentAddress\": {\n            \"type\": \"string\",\n            \"pattern\": \"^0x[a-fA-F0-9]{40}$\"\n          }\n        },\n        \"required\": [\n          \"workflowId\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"run_workflow\",\n      \"description\": \"Creates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\\r\\n\\r\\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"workflowId\": {\n            \"type\": \"string\"\n          },\n          \"sessionId\": {\n            \"type\": \"string\"\n          },\n          \"agentAddress\": {\n            \"type\": \"string\",\n            \"pattern\": \"^0x[a-fA-F0-9]{40}$\"\n          },\n          \"sessionToken\": {\n            \"type\": \"string\"\n          },\n          \"installedSkills\": {\n            \"type\": \"array\",\n            \"items\": {\n              \"type\": \"string\"\n            }\n          }\n        },\n        \"required\": [\n          \"workflowId\",\n          \"sessionId\",\n          \"agentAddress\",\n          \"sessionToken\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"get_workflow_status\",\n      \"description\": \"Reads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"runId\": {\n            \"type\": \"string\"\n          }\n        },\n        \"required\": [\n          \"runId\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    }\n  ],\n  \"operatingRules\": [],\n  \"requiredEnv\": {\n    \"FARMDASH_API_KEY\": {\n      \"description\": \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\",\n      \"required\": false\n    }\n  },\n  \"links\": {\n    \"website\": \"https://www.farmdash.one/\",\n    \"agentHub\": \"https://www.farmdash.one/agents\",\n    \"canonicalSkill\": \"https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md\",\n    \"documentation\": \"https://www.farmdash.one/docs\",\n    \"status\": \"https://www.farmdash.one/api/v1/agent/status\",\n    \"openApi\": \"https://www.farmdash.one/agents/openapi.yaml\",\n    \"mcpDiscovery\": \"https://www.farmdash.one/.well-known/mcp.json\",\n    \"fees\": \"https://www.farmdash.one/fees\",\n    \"security\": \"https://www.farmdash.one/security\"\n  }\n}\n\nArchive v0.1.9: 4 files, 11227 bytes\n\nFiles: skill-card.md (2575b), SKILL.json (5231b), SKILL.md (20423b), _meta.json (141b)\n\nFile v0.1.9:SKILL.md\n\n---\r\nname: FarmDash Trail Marshal\r\ndescription: \"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\"\r\ntags: [\"defi\",\"defi-workflows\",\"agent-orchestration\",\"workflow-automation\",\"multi-agent-system\",\"ai-agent-workflows\",\"policy-gated\",\"supervised-automation\",\"defi-playbooks\",\"quality-gates\",\"workflow-status\",\"session-management\",\"risk-management\",\"yield-farming-workflow\",\"airdrop-workflow\",\"zero-custody\",\"openclaw\",\"mcp\",\"web3\",\"farmdash\"]\r\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\r\nhomepage: https://www.farmdash.one/agents\r\nversion: \"1.1.2\"\r\nicon: 🪪\r\nenv:\r\n  FARMDASH_API_KEY:\r\n    description: \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\"\r\n    required: false\r\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"primaryEnv\":\"FARMDASH_API_KEY\",\"apiKeyRequired\":false,\"freeScoutKeyless\":true,\"freeScoutKey\":\"fd_scout_free\",\"execution\":\"guarded-orchestration-no-execution\"}}\r\n---\r\n\r\n# FarmDash Trail Marshal\r\n\r\n> Use this skill when running a multi-step DeFi goal: pick a named, confirmation-gated recipe instead of improvising the sequence at runtime.\r\n\r\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\r\n\r\n## What this skill does\r\n\r\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\r\n\r\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\r\n\r\n## Available tools\r\n\r\n### `list_workflows`\r\n\r\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\r\n\r\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\r\n\r\n**Returns shape:**\r\n\r\n```\r\n{\r\n  \"workflows\": [\r\n    {\r\n      \"id\": \"farm_hyperliquid\",\r\n      \"name\": \"Hyperliquid Point Farming\",\r\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\r\n      \"requiredTier\": \"syndicate\",\r\n      \"steps\": [\r\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\r\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\r\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\r\n      ]\r\n    }\r\n  ]\r\n}\r\n```\r\n\r\nThe detailed `steps` for each recipe are delivered at runtime as part of the `list_workflows` JSON response. Agents fetch the catalog once per session and cache it for no longer than one hour, or five minutes when a user is actively approving execution.\r\n\r\nCodebase alignment note: Wagon steps use `get_wallet_balances`; do not use the stale alias `get_balances`.\r\n\r\n**Agent posture:** Call `list_workflows` once at conversation start. Present recipes by name when the user describes a goal that matches one of them.\r\n\r\n### `plan_workflow`\r\n\r\nBuilds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\r\n\r\n- available steps.\r\n- missing steps.\r\n- state-changing steps.\r\n- confirmation count.\r\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\r\n\r\nUse this before telling the user a workflow is executable.\r\n\r\n### `run_workflow`\r\n\r\nCreates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\r\n\r\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\r\n\r\n### `get_workflow_status`\r\n\r\nReads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\r\n\r\n## Workflow catalog (high level)\r\n\r\n### Which Recipe for Which Goal\r\nIdle stables or ladder intent: `stablecoin_yield_ladder`, then `idle_capital_deploy`. Existing-book review: `rebalance_portfolio`, `rotate_quarterly`, or `yield_optimization`. Risk or exit: `protect_portfolio` for scans, `emergency_exit` for unwinds. Post-trade: `post_trade_ledger_review` then `post_execution_quality_review`. Pre-trade validation: `pre_trade_edge_audit`, `funding_carry_break_even_audit`, `perps_liquidation_buffer_check`. Continuous sessions: `session_command_center`. Never use `sybil_dilution` for transactions.\r\n\r\nEighteen recipes are published. Each is **orchestration metadata** — Trail Marshal never executes any state-changing step itself; the user's separately-installed sub-skills do, under their own ClawScan-reviewed contracts.\r\n\r\n| ID | Goal | Tier | User confirmations |\r\n|---|---|---|---|\r\n| `farm_hyperliquid` | Coordinate Hyperliquid points, spot exposure, and perps/funding research | Syndicate | per execution step |\r\n| `rotate_quarterly` | Compare current positions against fresh high-heat opportunities | Pioneer | per swap |\r\n| `protect_portfolio` | Risk-first scan and optional emergency mitigation | Pioneer | 0 if no action |\r\n| `tax_loss_harvest` | Identify loss-harvest candidates and quote realization paths | Pioneer | per swap |\r\n| `delta_neutral_setup` | Plan paired spot + perp exposure with non-atomic legging controls | Syndicate | research-only until a paired execution adapter exists |\r\n| `idle_capital_deploy` | Find idle stables/native assets and compare deployment targets | Scout catalog / Pioneer wallet data | per swap |\r\n| `sybil_dilution` | Compatibility ID for read-only Automation Compliance Review; no evasion transactions | Pioneer | 0 |\r\n| `yield_optimization` | Review harvest/redeposit candidates against Trail Heat | Pioneer | per swap |\r\n| `emergency_exit` | Plan a protocol exit and unwind route | Scout catalog / Pioneer wallet data | per swap |\r\n| `rebalance_portfolio` | Re-align portfolio to target weights and current Trail Heat | Pioneer | per swap |\r\n| `stablecoin_yield_ladder` | Build a risk-tiered stablecoin deployment ladder | Pioneer | per deposit intent |\r\n| `post_trade_ledger_review` | Reconcile activity and export records after execution | Pioneer | 0 |\r\n| `pre_trade_edge_audit` | Confirm positive net edge, clean route quality, fresh risk checks, and quote stability before execution | Pioneer | 0 before execution handoff |\r\n| `post_execution_quality_review` | Compare expected versus realized output before allowing dependent follow-on actions | Pioneer | 0 |\r\n| `perps_liquidation_buffer_check` | Reject or resize perps orders that do not survive normal volatility around liquidation | Pioneer | 0 before execution handoff |\r\n| `funding_carry_break_even_audit` | Validate funding trades after fees, slippage, margin pressure, and funding flip risk | Pioneer | 0 before execution handoff |\r\n| `session_command_center` | Coordinate context, event snapshots, workflow plans, and heartbeats | Pioneer | per execution handoff |\r\n| `hedged_airdrop_rotation` | Rotate into high-heat farms while planning hedge coverage | Syndicate | spot leg + hedge leg |\r\n\r\nEach recipe's full step graph is returned at runtime by `list_workflows`. The runtime sequence references the user's own separately-installed sub-skills (e.g. *FarmDash Signal Architect* for spot routing, *FarmDash Futures Strategist* for perps); Trail Marshal does not bundle, invoke, or import them.\r\n\r\n`sybil_dilution` is a legacy identifier only. Never generate transactions to imitate organic activity, dilute clustering signals, or evade protocol anti-abuse controls.\r\n\r\n## Permissions\r\n\r\nTrail Marshal can read the public workflow catalog without session state. `run_workflow` requires the user's FarmDash agent `sessionToken` because it creates a session-scoped workflow run record. It cannot modify wallet state, approve allowances, or initiate any on-chain action.\r\n\r\nIf a recipe in the catalog requires execution, the user's separately-installed sub-skill performs that work under its own ClawScan-reviewed contract — Trail Marshal stays out of the execution call path.\r\n\r\nIf the runtime catalog references a skill that is not installed in the user's environment, such as `farmdash-camp-guard`, `farmdash-supply-master`, `farmdash-hedge-warden`, or `farmdash-ledger-keeper`, mark that step as unavailable and continue with the safe subset. Never fabricate a missing tool call.\r\n\r\n## Tier model\r\n\r\n| Tier | Cost | Limits | Capability |\r\n|---|---|---|---|\r\n| **Scout** | Free | 5 req / 24h | Returns the public workflow catalog |\r\n| **Pioneer** | $39.99/mo | 1,500 req / day | Adds higher-rate catalog use; filtering is via the `filter` search string |\r\n| **Syndicate** | $199/mo | 50k req / day | Teams, serious agents, high-volume orchestration, webhooks, unrestricted CORS, advanced session/control tooling |\r\n\r\n## Composition patterns\r\n\r\n### Workflow Quality Gate (v0.2)\r\n\r\n### Gate Before You Present\r\nRun `plan_workflow` with `workflowId`, `installedSkills`, and `agentAddress` before calling any recipe executable. Executable requires every state-changing step's owning skill installed; a missing execution step forces `analysis_only`, a missing read-only step lowers confidence. Quote the confirmation count up front as per execution step unless the live catalog is stricter.\r\n\r\nBefore presenting a recipe as executable, classify every step:\r\n\r\n```json\r\n{\r\n  \"workflowId\": \"rotate_quarterly\",\r\n  \"availableSteps\": [],\r\n  \"missingSteps\": [],\r\n  \"stateChangingSteps\": [],\r\n  \"confirmationsRequired\": 0,\r\n  \"safeSubset\": true,\r\n  \"fallback\": \"research_only | analysis_only | halt\"\r\n}\r\n```\r\n\r\nRules:\r\n\r\n- A workflow is executable only when every state-changing step has its owning execution skill installed.\r\n- A missing read-only step lowers confidence; a missing execution step changes the recipe to `analysis_only`.\r\n- If the recipe includes `execute_swap` or `execute_perp_order`, quote the confirmation count as \"per execution step\" unless the live catalog supplies a stricter count.\r\n- If more than five minutes pass after the user reviews a plan, re-fetch `list_workflows` and re-run the read-only sense steps.\r\n- A catalog step is not evidence that the named tool supplies every field in its purpose. Validate live tool contracts and downgrade claims when a compatibility alias is narrower than its name.\r\n- Never treat submitted/broadcast as filled/confirmed, or a scalar score as a probability.\r\n- Multi-leg execution is non-atomic unless the adapter proves otherwise. Define leg order, maximum unhedged time, partial-fill handling, abort conditions, and unwind before the first signature.\r\n\r\n### The Sense / Decide / Present / Verify loop\r\n\r\nEvery Trail Marshal recipe respects this 4-phase pattern:\r\n\r\n```\r\nSENSE    → Read-only research and portfolio skills gather state\r\nDECIDE   → The recipe ranks options and presents a plan to the user\r\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\r\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle\r\n```\r\n\r\nIf a recipe skips SENSE, it is unsafe. If it skips VERIFY, the agent never improves. Trail Marshal recipes always include both.\r\n\r\n### Confirmation gate\r\n\r\n### Multi-Leg Execution Disclosure\r\nTell the user multi-leg execution is non-atomic unless the adapter proves otherwise, and state leg order, maximum unhedged time, partial-fill handling, abort conditions, and unwind before the first signature. Trail Marshal cannot pre-confirm anything; each confirmation happens in the owning sub-skill's quote path, and `delta_neutral_setup` stays research-only until a paired adapter exists.\r\n\r\nTrail Marshal cannot pre-confirm anything. Each user confirmation in a recipe happens at the moment the user's *separate* sub-skill presents its quote — not in Trail Marshal's call path. If asked to run a recipe without intermediate review, the agent should refuse and explain the contract.\r\n\r\nFor `delta_neutral_setup`, the current catalog intentionally stops before execution. Neither `execute_swap` nor `execute_perp_order` can atomically bind both legs, and standalone `funding_arb` execution is analysis-only. A future paired adapter must bind both legs, maximum basis/slippage, timeout, partial-fill policy, and unwind before this workflow becomes executable.\r\n\r\n### Quant workflow gate\r\n\r\nBefore presenting any recipe as action-ready, require objective/horizon, source timestamps and freshness, full cost stack, conservative net edge, downside/invalidation, portfolio concentration impact, and missing evidence. Unknown is not zero. Trail Heat, FarmScore, yield sustainability, sybil risk, and futures confidence use different methodologies and must never be multiplied into a synthetic probability.\r\n\r\nFail to `research_only` or `halt` when data is stale/degraded, a critical safety flag is unresolved, authoritative settlement is missing for a dependency, or an execution step cannot bind the parameters the user reviewed.\r\n\r\n### Tier composition\r\n\r\nA recipe's required tier is the **maximum** of its sub-skill tiers. When a user is one tier below a recipe's requirement, the agent should recognize the gap from `requiredTier`, offer the lower-tier subset of the recipe (e.g. research-only without the execution leg), and surface an upgrade link if the user wants the full version. Never silently downgrade a recipe without telling the user what is being skipped.\r\n\r\n## Reasoning guidelines for agents\r\n\r\n- **Speak in workflows, not tool soup.** Quote a recipe's name and goal when presenting a plan.\r\n- **Quote the `confirms` count up front** so the user knows how many user-confirmation steps the recipe involves.\r\n- **Re-fetch and re-present** if more than ~5 minutes pass between catalog read and the user's review.\r\n\r\n### Catalog Freshness Discipline\r\nCache `list_workflows` for no longer than one hour, or five minutes once the user is actively approving. If more than five minutes pass after the user reviews a plan, re-fetch the catalog, re-run the read-only Sense steps, and re-present before any handoff to an execution sub-skill.\r\n- **No new analysis.** Trail Marshal returns recipe metadata; it does not editorialize or override what other skills produce.\r\n\r\n## Risk warnings the agent should surface\r\n\r\nFor any recipe whose execution legs are owned by a separately-installed sub-skill, the agent should restate to the user that:\r\n\r\n- DeFi positions can lose value rapidly; airdrop rewards are speculative and not guaranteed.\r\n- Smart contracts can be exploited; even high-Trail-Heat protocols carry technical risk.\r\n- Cross-chain transfers are irreversible if the wrong address or chain is selected.\r\n- Slippage, gas, MEV, and routing fees materially affect realized returns.\r\n- Past Trail Heat performance does not guarantee future scores.\r\n- The user retains full responsibility for every on-chain decision they confirm via their own execution skill.\r\n\r\n## Commercial disclosure (inherited from sub-skills)\r\n\r\nWhen a recipe surfaces a `https://www.farmdash.one/go/{slug}` partnership route, the standard FarmDash commercial disclosure must be included in the same message: FarmDash may receive referral, affiliate, or routing compensation, and fee details live at [FarmDash Fee Structure](https://www.farmdash.one/fees). Trail Marshal does not introduce new partner routes of its own.\r\n\r\n## Versioning\r\n\r\n**v1.0 (current).** Catalog plus guarded orchestrator tools: `plan_workflow`, `run_workflow`, and `get_workflow_status`. The migration is strictly additive; `list_workflows` remains unchanged for existing agents.\r\n\r\n## Disclaimers\r\n\r\nWorkflow recipes are documentation, not financial advice. Any action performed by separately-installed sub-skills is the user's decision under that sub-skill's own contract. Trail Marshal performs no on-chain action; the user retains full responsibility for every decision they confirm.\r\n\r\n---\r\n\r\n**Install:** Copy this file into your OpenClaw workspace, or fetch `https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md`.\r\n\r\n**Companion skills (separately installed by the user):** FarmDash Trail Intelligence (research), FarmDash Wagon Steward (read-only portfolio), FarmDash Signal Architect (spot routing), FarmDash Futures Strategist (perps), FarmDash Camp Guard (risk), FarmDash Supply Master (yield), FarmDash Hedge Warden (hedges), FarmDash Ledger Keeper (records), FarmDash Autonomous Operator (session context).\r\n\r\n**Why FarmDash:** Unlike single-prompt agent chains, Trail Marshal runs named recipes with per-step confirmation counts, separates read-only steps from state-changing ones, and refuses to invent sequences at runtime.\r\n\r\n**FarmDash:** [Autonomous DeFi workflow intelligence](https://www.farmdash.one/)\r\n**Agent Hub:** [FarmDash autonomous DeFi workflow platform](https://www.farmdash.one/agents)\r\n**OpenAPI Spec:** [FarmDash API Schema](https://www.farmdash.one/agents/openapi.yaml)\r\n**MCP Config:** [FarmDash MCP Server](https://www.farmdash.one/.well-known/mcp.json)\r\n\r\n---\r\n\r\n## Optional Setup Check\r\n\r\nNo registration call is required to install this skill, browse workflow recipes, or use research-only planning. Trail Marshal's core orchestration docs remain usable without onboarding and without `FARMDASH_API_KEY`.\r\n\r\nOnly run the setup check if the user explicitly asks to verify FarmDash tier/setup status and agrees to send the listed metadata.\r\n\r\nData sent:\r\n\r\n- public `agentAddress`\r\n- `X-ClawHub-Skill: farmdash-trail-marshal`\r\n\r\nData not sent:\r\n\r\n- private keys, seed phrases, mnemonics, wallet exports, OAuth tokens, raw wallet secrets, or signed transactions\r\n\r\nOptional command after consent:\r\n\r\n```bash\r\ncurl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\r\n  -H \"Content-Type: application/json\" \\\r\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\r\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'\r\n```\r\n\r\nThis returns tier status and available tool access. Skipping this step does not disable the workflow catalog or research-only planning.\r\n\r\n### Next steps:\r\n1. Use the Scout workflow catalog without setup when no key is configured.\r\n2. Add `FARMDASH_API_KEY` only when the user wants higher-rate Pioneer/Syndicate workflow features.\r\n3. Browse the OpenAPI spec at `https://www.farmdash.one/agents/openapi.yaml`.\n\n<!-- farmdash-canonical-links:start -->\n\n## Official FarmDash Links\n\n- [FarmDash DeFi intelligence website](https://www.farmdash.one/)\n- [FarmDash Agent Hub](https://www.farmdash.one/agents)\n- [Canonical FarmDash Trail Marshal skill manual](https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md)\n- [Agent integration documentation](https://www.farmdash.one/docs)\n- [Live agent capability status](https://www.farmdash.one/api/v1/agent/status)\n- [OpenAPI contract](https://www.farmdash.one/agents/openapi.yaml)\n- [MCP discovery manifest](https://www.farmdash.one/.well-known/mcp.json)\n- [Fees and commercial terms](https://www.farmdash.one/fees)\n- [Security and authority boundaries](https://www.farmdash.one/security)\n\n<!-- farmdash-canonical-links:end -->\n\nFile v0.1.9:_meta.json\n\n{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.9\",\n  \"publishedAt\": 1788804353753\n}\n\nFile v0.1.9:skill-card.md\n\n## Description:\n\nUse when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\n\nThis skill is ready for commercial/non-commercial use.\n\n## Publisher:\n\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic)\n\n### License/Terms of Use:\n\nMIT-0\n\n## Use Case:\n\nExternal developers and DeFi agents use this skill to compose named FarmDash workflow recipes, plan quality-gated runs, inspect run status, and keep wallet-affecting actions behind separately installed execution skills with fresh user confirmation.\n\n### Deployment Geography for Use:\n\nGlobal\n\n## Known Risks and Mitigations:\n\nRisk: Users may mistake workflow planning or guarded run records for wallet execution authority.\n\nMitigation: Treat the skill as a planner only; require a separate execution skill, fresh data, and explicit user confirmation for any real transaction.\n\nRisk: Optional API-key or onboarding checks may send FarmDash a public agent address and session workflow metadata.\n\nMitigation: Use the optional key or setup check only after the user is comfortable sharing that metadata.\n\nRisk: A recipe may reference sub-skills that are not installed or cannot bind the parameters the user reviewed.\n\nMitigation: Downgrade to the safe subset, research-only, analysis-only, or halt instead of fabricating missing tool calls or skipping confirmation gates.\n\n## Reference(s):\n\n- [FarmDash Agent Hub](https://www.farmdash.one/agents)\n- [FarmDash API Schema](https://www.farmdash.one/agents/openapi.yaml)\n- [FarmDash MCP Server](https://www.farmdash.one/.well-known/mcp.json)\n- [FarmDash Documentation](https://www.farmdash.one/docs)\n- [Canonical FarmDash Trail Marshal Skill](https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md)\n\n## Skill Output:\n\n**Output Type(s):** [Text, Markdown, API calls, Shell commands, Configuration, Guidance]\n\n**Output Format:** [Markdown with JSON examples and optional shell commands]\n\n**Output Parameters:** [1D]\n\n**Other Properties Related to Output:** [Tool outputs include workflow catalogs, workflow plans, guarded run records, and run status; wallet execution remains outside this skill.]\n\n## Skill Version(s):\n\n0.1.9 (source: server release metadata; artifact frontmatter version 1.1.2)\n\n## Ethical Considerations:\n\nUsers should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment.\n\nFile v0.1.9:SKILL.json\n\n{\n  \"name\": \"FarmDash Trail Marshal\",\n  \"version\": \"1.1.2\",\n  \"description\": \"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\",\n  \"tags\": [\n    \"defi\",\n    \"defi-workflows\",\n    \"agent-orchestration\",\n    \"workflow-automation\",\n    \"multi-agent-system\",\n    \"ai-agent-workflows\",\n    \"policy-gated\",\n    \"supervised-automation\",\n    \"defi-playbooks\",\n    \"quality-gates\",\n    \"workflow-status\",\n    \"session-management\",\n    \"risk-management\",\n    \"yield-farming-workflow\",\n    \"airdrop-workflow\",\n    \"zero-custody\",\n    \"openclaw\",\n    \"mcp\",\n    \"web3\",\n    \"farmdash\"\n  ],\n  \"author\": \"FarmDash Pioneers (@Parmasanandgarlic)\",\n  \"homepage\": \"https://www.farmdash.one/agents\",\n  \"icon\": \"🪪\",\n  \"tools\": [\n    {\n      \"name\": \"list_workflows\",\n      \"description\": \"Returns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\\r\\n\\r\\n**\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"filter\": {\n            \"type\": \"string\"\n          }\n        },\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"plan_workflow\",\n      \"description\": \"Builds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\\r\\n\\r\\n- available steps.\\r\\n- missing steps.\\r\\n- state-changing steps.\\r\\n- confirmation count.\\r\\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\\r\\n\\r\\nUse this before telling the user a workflow is executable.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"workflowId\": {\n            \"type\": \"string\"\n          },\n          \"installedSkills\": {\n            \"type\": \"array\",\n            \"items\": {\n              \"type\": \"string\"\n            }\n          },\n          \"agentAddress\": {\n            \"type\": \"string\",\n            \"pattern\": \"^0x[a-fA-F0-9]{40}$\"\n          }\n        },\n        \"required\": [\n          \"workflowId\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"run_workflow\",\n      \"description\": \"Creates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\\r\\n\\r\\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"workflowId\": {\n            \"type\": \"string\"\n          },\n          \"sessionId\": {\n            \"type\": \"string\"\n          },\n          \"agentAddress\": {\n            \"type\": \"string\",\n            \"pattern\": \"^0x[a-fA-F0-9]{40}$\"\n          },\n          \"sessionToken\": {\n            \"type\": \"string\"\n          },\n          \"installedSkills\": {\n            \"type\": \"array\",\n            \"items\": {\n              \"type\": \"string\"\n            }\n          }\n        },\n        \"required\": [\n          \"workflowId\",\n          \"sessionId\",\n          \"agentAddress\",\n          \"sessionToken\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"get_workflow_status\",\n      \"description\": \"Reads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"runId\": {\n            \"type\": \"string\"\n          }\n        },\n        \"required\": [\n          \"runId\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    }\n  ],\n  \"operatingRules\": [],\n  \"requiredEnv\": {\n    \"FARMDASH_API_KEY\": {\n      \"description\": \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\",\n      \"required\": false\n    }\n  },\n  \"links\": {\n    \"website\": \"https://www.farmdash.one/\",\n    \"agentHub\": \"https://www.farmdash.one/agents\",\n    \"canonicalSkill\": \"https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md\",\n    \"documentation\": \"https://www.farmdash.one/docs\",\n    \"status\": \"https://www.farmdash.one/api/v1/agent/status\",\n    \"openApi\": \"https://www.farmdash.one/agents/openapi.yaml\",\n    \"mcpDiscovery\": \"https://www.farmdash.one/.well-known/mcp.json\",\n    \"fees\": \"https://www.farmdash.one/fees\",\n    \"security\": \"https://www.farmdash.one/security\"\n  }\n}\n\nArchive v0.1.8: 4 files, 10114 bytes\n\nFiles: skill-card.md (2800b), SKILL.json (3013b), SKILL.md (17482b), _meta.json (141b)\n\nFile v0.1.8:SKILL.md\n\n---\r\nname: FarmDash Trail Marshal\r\ndescription: \"Orchestrate guarded multi-skill DeFi workflows with named recipes, quality gates, session-scoped runs, status tracking, and no key custody.\"\ntags: [\"defi\", \"defi-workflows\", \"agent-orchestration\", \"workflow-automation\", \"multi-agent-system\", \"ai-agent-workflows\", \"defi-automation\", \"quality-gates\", \"workflow-status\", \"session-management\", \"risk-management\", \"yield-farming-workflow\", \"airdrop-workflow\", \"zero-custody\", \"openclaw\", \"mcp\", \"farmdash\"]\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\r\nhomepage: https://www.farmdash.one/agents\r\nversion: \"1.1.1\"\nicon: 🪪\r\nenv:\r\n  FARMDASH_API_KEY:\r\n    description: \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\"\r\n    required: false\r\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"primaryEnv\":\"FARMDASH_API_KEY\",\"apiKeyRequired\":false,\"freeScoutKeyless\":true,\"freeScoutKey\":\"fd_scout_free\",\"execution\":\"guarded-orchestration-no-execution\"}}\r\n---\r\n\r\n# FarmDash Trail Marshal\r\n\r\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\r\n\r\n## What this skill does\r\n\r\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\r\n\r\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\r\n\r\n## Available tools\r\n\r\n### `list_workflows`\r\n\r\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\r\n\r\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\r\n\r\n**Returns shape:**\r\n\r\n```\r\n{\r\n  \"workflows\": [\r\n    {\r\n      \"id\": \"farm_hyperliquid\",\r\n      \"name\": \"Hyperliquid Point Farming\",\r\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\r\n      \"requiredTier\": \"syndicate\",\r\n      \"steps\": [\r\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\r\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\r\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\r\n      ]\r\n    }\r\n  ]\r\n}\r\n```\r\n\r\nThe detailed `steps` for each recipe are delivered at runtime as part of the `list_workflows` JSON response. Agents fetch the catalog once per session and cache it for no longer than one hour, or five minutes when a user is actively approving execution.\r\n\r\nCodebase alignment note: Wagon steps use `get_wallet_balances`; do not use the stale alias `get_balances`.\r\n\r\n**Agent posture:** Call `list_workflows` once at conversation start. Present recipes by name when the user describes a goal that matches one of them.\r\n\r\n### `plan_workflow`\r\n\r\nBuilds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\r\n\r\n- available steps.\r\n- missing steps.\r\n- state-changing steps.\r\n- confirmation count.\r\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\r\n\r\nUse this before telling the user a workflow is executable.\r\n\r\n### `run_workflow`\r\n\r\nCreates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\r\n\r\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\r\n\r\n### `get_workflow_status`\r\n\r\nReads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\r\n\r\n## Workflow catalog (high level)\r\n\r\nEighteen recipes are published. Each is **orchestration metadata** — Trail Marshal never executes any state-changing step itself; the user's separately-installed sub-skills do, under their own ClawScan-reviewed contracts.\r\n\r\n| ID | Goal | Tier | User confirmations |\r\n|---|---|---|---|\r\n| `farm_hyperliquid` | Coordinate Hyperliquid points, spot exposure, and perps/funding research | Syndicate | per execution step |\r\n| `rotate_quarterly` | Compare current positions against fresh high-heat opportunities | Pioneer | per swap |\r\n| `protect_portfolio` | Risk-first scan and optional emergency mitigation | Pioneer | 0 if no action |\r\n| `tax_loss_harvest` | Identify loss-harvest candidates and quote realization paths | Pioneer | per swap |\r\n| `delta_neutral_setup` | Plan paired spot + perp exposure with non-atomic legging controls | Syndicate | research-only until a paired execution adapter exists |\n| `idle_capital_deploy` | Find idle stables/native assets and compare deployment targets | Scout catalog / Pioneer wallet data | per swap |\r\n| `sybil_dilution` | Compatibility ID for read-only Automation Compliance Review; no evasion transactions | Pioneer | 0 |\n| `yield_optimization` | Review harvest/redeposit candidates against Trail Heat | Pioneer | per swap |\r\n| `emergency_exit` | Plan a protocol exit and unwind route | Scout catalog / Pioneer wallet data | per swap |\r\n| `rebalance_portfolio` | Re-align portfolio to target weights and current Trail Heat | Pioneer | per swap |\r\n| `stablecoin_yield_ladder` | Build a risk-tiered stablecoin deployment ladder | Pioneer | per deposit intent |\r\n| `post_trade_ledger_review` | Reconcile activity and export records after execution | Pioneer | 0 |\r\n| `pre_trade_edge_audit` | Confirm positive net edge, clean route quality, fresh risk checks, and quote stability before execution | Pioneer | 0 before execution handoff |\r\n| `post_execution_quality_review` | Compare expected versus realized output before allowing dependent follow-on actions | Pioneer | 0 |\r\n| `perps_liquidation_buffer_check` | Reject or resize perps orders that do not survive normal volatility around liquidation | Pioneer | 0 before execution handoff |\r\n| `funding_carry_break_even_audit` | Validate funding trades after fees, slippage, margin pressure, and funding flip risk | Pioneer | 0 before execution handoff |\r\n| `session_command_center` | Coordinate context, event snapshots, workflow plans, and heartbeats | Pioneer | per execution handoff |\r\n| `hedged_airdrop_rotation` | Rotate into high-heat farms while planning hedge coverage | Syndicate | spot leg + hedge leg |\r\n\r\nEach recipe's full step graph is returned at runtime by `list_workflows`. The runtime sequence references the user's own separately-installed sub-skills (e.g. *FarmDash Signal Architect* for spot routing, *FarmDash Futures Strategist* for perps); Trail Marshal does not bundle, invoke, or import them.\n\n`sybil_dilution` is a legacy identifier only. Never generate transactions to imitate organic activity, dilute clustering signals, or evade protocol anti-abuse controls.\n\r\n## Permissions\r\n\r\nTrail Marshal can read the public workflow catalog without session state. `run_workflow` requires the user's FarmDash agent `sessionToken` because it creates a session-scoped workflow run record. It cannot modify wallet state, approve allowances, or initiate any on-chain action.\r\n\r\nIf a recipe in the catalog requires execution, the user's separately-installed sub-skill performs that work under its own ClawScan-reviewed contract — Trail Marshal stays out of the execution call path.\r\n\r\nIf the runtime catalog references a skill that is not installed in the user's environment, such as `farmdash-camp-guard`, `farmdash-supply-master`, `farmdash-hedge-warden`, or `farmdash-ledger-keeper`, mark that step as unavailable and continue with the safe subset. Never fabricate a missing tool call.\r\n\r\n## Tier model\r\n\r\n| Tier | Cost | Limits | Capability |\r\n|---|---|---|---|\r\n| **Scout** | Free | 5 req / 24h | Returns the public workflow catalog |\r\n| **Pioneer** | $39.99/mo | 500 req / day | Adds higher-rate catalog use; filtering is via the `filter` search string |\r\n| **Syndicate** | $199/mo | 50k req / day | Teams, serious agents, high-volume orchestration, webhooks, unrestricted CORS, advanced session/control tooling |\r\n\r\n## Composition patterns\r\n\r\n### Workflow Quality Gate (v0.2)\r\n\r\nBefore presenting a recipe as executable, classify every step:\r\n\r\n```json\r\n{\r\n  \"workflowId\": \"rotate_quarterly\",\r\n  \"availableSteps\": [],\r\n  \"missingSteps\": [],\r\n  \"stateChangingSteps\": [],\r\n  \"confirmationsRequired\": 0,\r\n  \"safeSubset\": true,\r\n  \"fallback\": \"research_only | analysis_only | halt\"\r\n}\r\n```\r\n\r\nRules:\n\r\n- A workflow is executable only when every state-changing step has its owning execution skill installed.\r\n- A missing read-only step lowers confidence; a missing execution step changes the recipe to `analysis_only`.\r\n- If the recipe includes `execute_swap` or `execute_perp_order`, quote the confirmation count as \"per execution step\" unless the live catalog supplies a stricter count.\r\n- If more than five minutes pass after the user reviews a plan, re-fetch `list_workflows` and re-run the read-only sense steps.\n- A catalog step is not evidence that the named tool supplies every field in its purpose. Validate live tool contracts and downgrade claims when a compatibility alias is narrower than its name.\n- Never treat submitted/broadcast as filled/confirmed, or a scalar score as a probability.\n- Multi-leg execution is non-atomic unless the adapter proves otherwise. Define leg order, maximum unhedged time, partial-fill handling, abort conditions, and unwind before the first signature.\n\r\n### The Sense / Decide / Present / Verify loop\r\n\r\nEvery Trail Marshal recipe respects this 4-phase pattern:\r\n\r\n```\r\nSENSE    → Read-only research and portfolio skills gather state\r\nDECIDE   → The recipe ranks options and presents a plan to the user\r\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\r\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle\r\n```\r\n\r\nIf a recipe skips SENSE, it is unsafe. If it skips VERIFY, the agent never improves. Trail Marshal recipes always include both.\r\n\r\n### Confirmation gate\n\r\nTrail Marshal cannot pre-confirm anything. Each user confirmation in a recipe happens at the moment the user's *separate* sub-skill presents its quote — not in Trail Marshal's call path. If asked to run a recipe without intermediate review, the agent should refuse and explain the contract.\n\nFor `delta_neutral_setup`, the current catalog intentionally stops before execution. Neither `execute_swap` nor `execute_perp_order` can atomically bind both legs, and standalone `funding_arb` execution is analysis-only. A future paired adapter must bind both legs, maximum basis/slippage, timeout, partial-fill policy, and unwind before this workflow becomes executable.\n\n### Quant workflow gate\n\nBefore presenting any recipe as action-ready, require objective/horizon, source timestamps and freshness, full cost stack, conservative net edge, downside/invalidation, portfolio concentration impact, and missing evidence. Unknown is not zero. Trail Heat, FarmScore, yield sustainability, sybil risk, and futures confidence use different methodologies and must never be multiplied into a synthetic probability.\n\nFail to `research_only` or `halt` when data is stale/degraded, a critical safety flag is unresolved, authoritative settlement is missing for a dependency, or an execution step cannot bind the parameters the user reviewed.\n\r\n### Tier composition\r\n\r\nA recipe's required tier is the **maximum** of its sub-skill tiers. When a user is one tier below a recipe's requirement, the agent should recognize the gap from `requiredTier`, offer the lower-tier subset of the recipe (e.g. research-only without the execution leg), and surface an upgrade link if the user wants the full version. Never silently downgrade a recipe without telling the user what is being skipped.\r\n\r\n## Reasoning guidelines for agents\r\n\r\n- **Speak in workflows, not tool soup.** Quote a recipe's name and goal when presenting a plan.\r\n- **Quote the `confirms` count up front** so the user knows how many user-confirmation steps the recipe involves.\r\n- **Re-fetch and re-present** if more than ~5 minutes pass between catalog read and the user's review.\r\n- **No new analysis.** Trail Marshal returns recipe metadata; it does not editorialize or override what other skills produce.\r\n\r\n## Risk warnings the agent should surface\r\n\r\nFor any recipe whose execution legs are owned by a separately-installed sub-skill, the agent should restate to the user that:\r\n\r\n- DeFi positions can lose value rapidly; airdrop rewards are speculative and not guaranteed.\r\n- Smart contracts can be exploited; even high-Trail-Heat protocols carry technical risk.\r\n- Cross-chain transfers are irreversible if the wrong address or chain is selected.\r\n- Slippage, gas, MEV, and routing fees materially affect realized returns.\r\n- Past Trail Heat performance does not guarantee future scores.\r\n- The user retains full responsibility for every on-chain decision they confirm via their own execution skill.\r\n\r\n## Commercial disclosure (inherited from sub-skills)\r\n\r\nWhen a recipe surfaces a `https://www.farmdash.one/go/{slug}` partnership route, the standard FarmDash commercial disclosure must be included in the same message: FarmDash may receive referral, affiliate, or routing compensation, and fee details live at [FarmDash Fee Structure](https://www.farmdash.one/fees). Trail Marshal does not introduce new partner routes of its own.\r\n\r\n## Versioning\r\n\r\n**v1.0 (current).** Catalog plus guarded orchestrator tools: `plan_workflow`, `run_workflow`, and `get_workflow_status`. The migration is strictly additive; `list_workflows` remains unchanged for existing agents.\r\n\r\n## Disclaimers\r\n\r\nWorkflow recipes are documentation, not financial advice. Any action performed by separately-installed sub-skills is the user's decision under that sub-skill's own contract. Trail Marshal performs no on-chain action; the user retains full responsibility for every decision they confirm.\r\n\r\n---\r\n\r\n**Install:** Copy this file into your OpenClaw workspace, or fetch `https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md`.\r\n\r\n**Companion skills (separately installed by the user):** FarmDash Trail Intelligence (research), FarmDash Wagon Steward (read-only portfolio), FarmDash Signal Architect (spot routing), FarmDash Futures Strategist (perps), FarmDash Camp Guard (risk), FarmDash Supply Master (yield), FarmDash Hedge Warden (hedges), FarmDash Ledger Keeper (records), FarmDash Autonomous Operator (session context).\r\n\r\n**FarmDash:** [Autonomous DeFi workflow intelligence](https://www.farmdash.one/)\n**Agent Hub:** [FarmDash autonomous DeFi workflow platform](https://www.farmdash.one/agents)\n**OpenAPI Spec:** [FarmDash API Schema](https://www.farmdash.one/agents/openapi.yaml)\r\n**MCP Config:** [FarmDash MCP Server](https://www.farmdash.one/.well-known/mcp.json)\r\n\r\n---\r\n\r\n## Optional Setup Check\r\n\r\nNo registration call is required to install this skill, browse workflow recipes, or use research-only planning. Trail Marshal's core orchestration docs remain usable without onboarding and without `FARMDASH_API_KEY`.\r\n\r\nOnly run the setup check if the user explicitly asks to verify FarmDash tier/setup status and agrees to send the listed metadata.\r\n\r\nData sent:\r\n\r\n- public `agentAddress`\r\n- `X-ClawHub-Skill: farmdash-trail-marshal`\r\n\r\nData not sent:\r\n\r\n- private keys, seed phrases, mnemonics, wallet exports, OAuth tokens, raw wallet secrets, or signed transactions\r\n\r\nOptional command after consent:\r\n\r\n```bash\r\ncurl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\r\n  -H \"Content-Type: application/json\" \\\r\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\r\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'\r\n```\r\n\r\nThis returns tier status and available tool access. Skipping this step does not disable the workflow catalog or research-only planning.\r\n\r\n### Next steps:\r\n1. Use the Scout workflow catalog without setup when no key is configured.\r\n2. Add `FARMDASH_API_KEY` only when the user wants higher-rate Pioneer/Syndicate workflow features.\r\n3. Browse the OpenAPI spec at `https://www.farmdash.one/agents/openapi.yaml`.\n\nFile v0.1.8:_meta.json\n\n{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.8\",\n  \"publishedAt\": 1785819741262\n}\n\nFile v0.1.8:skill-card.md\n\n## Description: <br>\nOrchestrate guarded multi-skill DeFi workflows with named recipes, quality gates, session-scoped runs, status tracking, and no key custody. <br>\n\nThis skill is ready for commercial/non-commercial use. <br>\n\n## Publisher: <br>\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic) <br>\n\n### License/Terms of Use: <br>\nMIT-0 <br>\n\n\n## Use Case: <br>\nExternal developers and OpenClaw agent users use this skill to plan, quality-gate, and track multi-skill DeFi workflows while leaving swaps, deposits, perps orders, and other state-changing actions to separately installed execution skills. <br>\n\n### Deployment Geography for Use: <br>\nGlobal <br>\n\n## Known Risks and Mitigations: <br>\nRisk: A planned DeFi workflow may include execution steps handled by separately installed skills. <br>\nMitigation: Require each execution skill to present fresh data and obtain explicit user confirmation under its own permission and confirmation flow. <br>\nRisk: Workflow runs may send an agent address, session token, or optional FarmDash API key for session or paid-tier features. <br>\nMitigation: Share only the minimum required session or tier metadata, keep private keys and wallet secrets out of all calls, and treat FARMDASH_API_KEY as optional unless higher-rate features are needed. <br>\nRisk: Missing companion skills can make a recipe incomplete or analysis-only. <br>\nMitigation: Classify unavailable steps before presenting a workflow as executable and continue only with the safe subset when required sub-skills are absent. <br>\n\n\n## Reference(s): <br>\n- [ClawHub Skill Page](https://clawhub.ai/parmasanandgarlic/skills/farmdash-trail-marshal) <br>\n- [FarmDash Agent Hub](https://www.farmdash.one/agents) <br>\n- [FarmDash API Schema](https://www.farmdash.one/agents/openapi.yaml) <br>\n- [FarmDash MCP Server](https://www.farmdash.one/.well-known/mcp.json) <br>\n- [FarmDash Trail Marshal Skill Source](https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md) <br>\n\n\n## Skill Output: <br>\n**Output Type(s):** [text, markdown, shell commands, configuration, guidance] <br>\n**Output Format:** [Markdown guidance with JSON-shaped workflow catalog, plan, run, and status data.] <br>\n**Output Parameters:** [1D] <br>\n**Other Properties Related to Output:** [Produces workflow planning and status information only; state-changing DeFi execution remains outside this skill.] <br>\n\n## Skill Version(s): <br>\n0.1.8 (source: ClawHub release metadata; artifact frontmatter reports 1.1.1) <br>\n\n## Ethical Considerations: <br>\nUsers should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment. <br>\n\nFile v0.1.8:SKILL.json\n\n{\n  \"name\": \"FarmDash Trail Marshal\",\n  \"version\": \"1.1.1\",\n  \"description\": \"Orchestrate guarded multi-skill DeFi workflows with named recipes, quality gates, session-scoped runs, status tracking, and no key custody.\",\n  \"tags\": [\n    \"defi\",\n    \"defi-workflows\",\n    \"agent-orchestration\",\n    \"workflow-automation\",\n    \"multi-agent-system\",\n    \"ai-agent-workflows\",\n    \"defi-automation\",\n    \"quality-gates\",\n    \"workflow-status\",\n    \"session-management\",\n    \"risk-management\",\n    \"yield-farming-workflow\",\n    \"airdrop-workflow\",\n    \"zero-custody\",\n    \"openclaw\",\n    \"mcp\",\n    \"farmdash\"\n  ],\n  \"author\": \"FarmDash Pioneers (@Parmasanandgarlic)\",\n  \"homepage\": \"https://www.farmdash.one/agents\",\n  \"icon\": \"🪪\",\n  \"tools\": [\n    {\n      \"name\": \"list_workflows\",\n      \"description\": \"Returns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\\r\\n\\r\\n**\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"plan_workflow\",\n      \"description\": \"Builds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\\r\\n\\r\\n- available steps.\\r\\n- missing steps.\\r\\n- state-changing steps.\\r\\n- confirmation count.\\r\\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\\r\\n\\r\\nUse this before telling the user a workflow is executable.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"run_workflow\",\n      \"description\": \"Creates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\\r\\n\\r\\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"get_workflow_status\",\n      \"description\": \"Reads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    }\n  ],\n  \"operatingRules\": [],\n  \"requiredEnv\": {\n    \"FARMDASH_API_KEY\": {\n      \"description\": \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\",\n      \"required\": false\n    }\n  }\n}\n\nArchive v0.1.7: 4 files, 9262 bytes\n\nFiles: skill-card.md (2397b), SKILL.json (3237b), SKILL.md (15627b), _meta.json (141b)\n\nFile v0.1.7:SKILL.md\n\n---\nname: FarmDash Trail Marshal\ndescription: \"Guarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action — every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate.\"\ntags: [\"defi\", \"ai-agent\", \"autonomous-agent\", \"openclaw\", \"clawhub\", \"mcp\", \"crypto\", \"web3\", \"onchain\", \"orchestration\", \"workflow\", \"agent-workflows\", \"multi-agent\", \"quality-gates\", \"risk-management\", \"yield-farming\", \"airdrop\", \"read-only\", \"zero-custody\", \"farmdash\", \"trail-marshal\"]\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\nhomepage: https://www.farmdash.one/agents\nversion: \"1.0.1\"\nicon: 🪪\nenv:\n  FARMDASH_API_KEY:\n    description: \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\"\n    required: false\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"primaryEnv\":\"FARMDASH_API_KEY\",\"apiKeyRequired\":false,\"freeScoutKeyless\":true,\"freeScoutKey\":\"fd_scout_free\",\"execution\":\"guarded-orchestration-no-execution\"}}\n---\n\n# FarmDash Trail Marshal\n\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\n\n## What this skill does\n\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\n\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\n\n## Available tools\n\n### `list_workflows`\n\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\n\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\n\n**Returns shape:**\n\n```\n{\n  \"workflows\": [\n    {\n      \"id\": \"farm_hyperliquid\",\n      \"name\": \"Hyperliquid Point Farming\",\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\n      \"requiredTier\": \"syndicate\",\n      \"steps\": [\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\n      ]\n    }\n  ]\n}\n```\n\nThe detailed `steps` for each recipe are delivered at runtime as part of the `list_workflows` JSON response. Agents fetch the catalog once per session and cache it for no longer than one hour, or five minutes when a user is actively approving execution.\n\nCodebase alignment note: Wagon steps use `get_wallet_balances`; do not use the stale alias `get_balances`.\n\n**Agent posture:** Call `list_workflows` once at conversation start. Present recipes by name when the user describes a goal that matches one of them.\n\n### `plan_workflow`\n\nBuilds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\n\n- available steps.\n- missing steps.\n- state-changing steps.\n- confirmation count.\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\n\nUse this before telling the user a workflow is executable.\n\n### `run_workflow`\n\nCreates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\n\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\n\n### `get_workflow_status`\n\nReads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\n\n## Workflow catalog (high level)\n\nEighteen recipes are published. Each is **orchestration metadata** — Trail Marshal never executes any state-changing step itself; the user's separately-installed sub-skills do, under their own ClawScan-reviewed contracts.\n\n| ID | Goal | Tier | User confirmations |\n|---|---|---|---|\n| `farm_hyperliquid` | Coordinate Hyperliquid points, spot exposure, and perps/funding research | Syndicate | per execution step |\n| `rotate_quarterly` | Compare current positions against fresh high-heat opportunities | Pioneer | per swap |\n| `protect_portfolio` | Risk-first scan and optional emergency mitigation | Pioneer | 0 if no action |\n| `tax_loss_harvest` | Identify loss-harvest candidates and quote realization paths | Pioneer | per swap |\n| `delta_neutral_setup` | Coordinate paired spot + perp exposure | Syndicate | spot leg + perps leg |\n| `idle_capital_deploy` | Find idle stables/native assets and compare deployment targets | Scout catalog / Pioneer wallet data | per swap |\n| `sybil_dilution` | Review sybil-risk clusters and propose lower-correlation behavior | Pioneer | per action |\n| `yield_optimization` | Review harvest/redeposit candidates against Trail Heat | Pioneer | per swap |\n| `emergency_exit` | Plan a protocol exit and unwind route | Scout catalog / Pioneer wallet data | per swap |\n| `rebalance_portfolio` | Re-align portfolio to target weights and current Trail Heat | Pioneer | per swap |\n| `stablecoin_yield_ladder` | Build a risk-tiered stablecoin deployment ladder | Pioneer | per deposit intent |\n| `post_trade_ledger_review` | Reconcile activity and export records after execution | Pioneer | 0 |\n| `pre_trade_edge_audit` | Confirm positive net edge, clean route quality, fresh risk checks, and quote stability before execution | Pioneer | 0 before execution handoff |\n| `post_execution_quality_review` | Compare expected versus realized output before allowing dependent follow-on actions | Pioneer | 0 |\n| `perps_liquidation_buffer_check` | Reject or resize perps orders that do not survive normal volatility around liquidation | Pioneer | 0 before execution handoff |\n| `funding_carry_break_even_audit` | Validate funding trades after fees, slippage, margin pressure, and funding flip risk | Pioneer | 0 before execution handoff |\n| `session_command_center` | Coordinate context, event snapshots, workflow plans, and heartbeats | Pioneer | per execution handoff |\n| `hedged_airdrop_rotation` | Rotate into high-heat farms while planning hedge coverage | Syndicate | spot leg + hedge leg |\n\nEach recipe's full step graph is returned at runtime by `list_workflows`. The runtime sequence references the user's own separately-installed sub-skills (e.g. *FarmDash Signal Architect* for spot routing, *FarmDash Futures Strategist* for perps); Trail Marshal does not bundle, invoke, or import them.\n\n## Permissions\n\nTrail Marshal can read the public workflow catalog without session state. `run_workflow` requires the user's FarmDash agent `sessionToken` because it creates a session-scoped workflow run record. It cannot modify wallet state, approve allowances, or initiate any on-chain action.\n\nIf a recipe in the catalog requires execution, the user's separately-installed sub-skill performs that work under its own ClawScan-reviewed contract — Trail Marshal stays out of the execution call path.\n\nIf the runtime catalog references a skill that is not installed in the user's environment, such as `farmdash-camp-guard`, `farmdash-supply-master`, `farmdash-hedge-warden`, or `farmdash-ledger-keeper`, mark that step as unavailable and continue with the safe subset. Never fabricate a missing tool call.\n\n## Tier model\n\n| Tier | Cost | Limits | Capability |\n|---|---|---|---|\n| **Scout** | Free | 5 req / 24h | Returns the public workflow catalog |\n| **Pioneer** | $39.99/mo | 500 req / day | Adds higher-rate catalog use; filtering is via the `filter` search string |\n| **Syndicate** | $199/mo | 50k req / day | Teams, serious agents, high-volume orchestration, webhooks, unrestricted CORS, advanced session/control tooling |\n\n## Composition patterns\n\n### Workflow Quality Gate (v0.2)\n\nBefore presenting a recipe as executable, classify every step:\n\n```json\n{\n  \"workflowId\": \"rotate_quarterly\",\n  \"availableSteps\": [],\n  \"missingSteps\": [],\n  \"stateChangingSteps\": [],\n  \"confirmationsRequired\": 0,\n  \"safeSubset\": true,\n  \"fallback\": \"research_only | analysis_only | halt\"\n}\n```\n\nRules:\n\n- A workflow is executable only when every state-changing step has its owning execution skill installed.\n- A missing read-only step lowers confidence; a missing execution step changes the recipe to `analysis_only`.\n- If the recipe includes `execute_swap` or `execute_perp_order`, quote the confirmation count as \"per execution step\" unless the live catalog supplies a stricter count.\n- If more than five minutes pass after the user reviews a plan, re-fetch `list_workflows` and re-run the read-only sense steps.\n\n### The Sense / Decide / Present / Verify loop\n\nEvery Trail Marshal recipe respects this 4-phase pattern:\n\n```\nSENSE    → Read-only research and portfolio skills gather state\nDECIDE   → The recipe ranks options and presents a plan to the user\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle\n```\n\nIf a recipe skips SENSE, it is unsafe. If it skips VERIFY, the agent never improves. Trail Marshal recipes always include both.\n\n### Confirmation gate\n\nTrail Marshal cannot pre-confirm anything. Each user confirmation in a recipe happens at the moment the user's *separate* sub-skill presents its quote — not in Trail Marshal's call path. If asked to run a recipe without intermediate review, the agent should refuse and explain the contract.\n\n### Tier composition\n\nA recipe's required tier is the **maximum** of its sub-skill tiers. When a user is one tier below a recipe's requirement, the agent should recognize the gap from `requiredTier`, offer the lower-tier subset of the recipe (e.g. research-only without the execution leg), and surface an upgrade link if the user wants the full version. Never silently downgrade a recipe without telling the user what is being skipped.\n\n## Reasoning guidelines for agents\n\n- **Speak in workflows, not tool soup.** Quote a recipe's name and goal when presenting a plan.\n- **Quote the `confirms` count up front** so the user knows how many user-confirmation steps the recipe involves.\n- **Re-fetch and re-present** if more than ~5 minutes pass between catalog read and the user's review.\n- **No new analysis.** Trail Marshal returns recipe metadata; it does not editorialize or override what other skills produce.\n\n## Risk warnings the agent should surface\n\nFor any recipe whose execution legs are owned by a separately-installed sub-skill, the agent should restate to the user that:\n\n- DeFi positions can lose value rapidly; airdrop rewards are speculative and not guaranteed.\n- Smart contracts can be exploited; even high-Trail-Heat protocols carry technical risk.\n- Cross-chain transfers are irreversible if the wrong address or chain is selected.\n- Slippage, gas, MEV, and routing fees materially affect realized returns.\n- Past Trail Heat performance does not guarantee future scores.\n- The user retains full responsibility for every on-chain decision they confirm via their own execution skill.\n\n## Commercial disclosure (inherited from sub-skills)\n\nWhen a recipe surfaces a `https://www.farmdash.one/go/{slug}` partnership route, the standard FarmDash commercial disclosure must be included in the same message: FarmDash may receive referral, affiliate, or routing compensation, and fee details live at [FarmDash Fee Structure](https://www.farmdash.one/fees). Trail Marshal does not introduce new partner routes of its own.\n\n## Versioning\n\n**v1.0 (current).** Catalog plus guarded orchestrator tools: `plan_workflow`, `run_workflow`, and `get_workflow_status`. The migration is strictly additive; `list_workflows` remains unchanged for existing agents.\n\n## Disclaimers\n\nWorkflow recipes are documentation, not financial advice. Any action performed by separately-installed sub-skills is the user's decision under that sub-skill's own contract. Trail Marshal performs no on-chain action; the user retains full responsibility for every decision they confirm.\n\n---\n\n**Install:** Copy this file into your OpenClaw workspace, or fetch `https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md`.\n\n**Companion skills (separately installed by the user):** FarmDash Trail Intelligence (research), FarmDash Wagon Steward (read-only portfolio), FarmDash Signal Architect (spot routing), FarmDash Futures Strategist (perps), FarmDash Camp Guard (risk), FarmDash Supply Master (yield), FarmDash Hedge Warden (hedges), FarmDash Ledger Keeper (records), FarmDash Autonomous Operator (session context).\n\n**Agent Hub:** [FarmDash Agentic OS](https://www.farmdash.one/agents)\n**OpenAPI Spec:** [FarmDash API Schema](https://www.farmdash.one/agents/openapi.yaml)\n**MCP Config:** [FarmDash MCP Server](https://www.farmdash.one/.well-known/mcp.json)\n\n---\n\n## Optional Setup Check\n\nNo registration call is required to install this skill, browse workflow recipes, or use research-only planning. Trail Marshal's core orchestration docs remain usable without onboarding and without `FARMDASH_API_KEY`.\n\nOnly run the setup check if the user explicitly asks to verify FarmDash tier/setup status and agrees to send the listed metadata.\n\nData sent:\n\n- public `agentAddress`\n- `X-ClawHub-Skill: farmdash-trail-marshal`\n\nData not sent:\n\n- private keys, seed phrases, mnemonics, wallet exports, OAuth tokens, raw wallet secrets, or signed transactions\n\nOptional command after consent:\n\n```bash\ncurl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\n  -H \"Content-Type: application/json\" \\\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'\n```\n\nThis returns tier status and available tool access. Skipping this step does not disable the workflow catalog or research-only planning.\n\n### Next steps:\n1. Use the Scout workflow catalog without setup when no key is configured.\n2. Add `FARMDASH_API_KEY` only when the user wants higher-rate Pioneer/Syndicate workflow features.\n3. Browse the OpenAPI spec at `https://www.farmdash.one/agents/openapi.yaml`.\n\nFile v0.1.7:_meta.json\n\n{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.7\",\n  \"publishedAt\": 1782610493625\n}\n\nFile v0.1.7:skill-card.md\n\n## Description: <br>\nFarmDash Trail Marshal helps agents plan guarded DeFi workflows by listing recipes, building quality gates, recording session workflow runs, and reporting status without executing wallet or trading actions. <br>\n\nThis skill is ready for commercial/non-commercial use. <br>\n\n## Publisher: <br>\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic) <br>\n\n### License/Terms of Use: <br>\nMIT-0 <br>\n\n\n## Use Case: <br>\nDevelopers and agent operators use this skill to coordinate DeFi workflow recipes, check installed companion skills, and keep state-changing actions behind separate confirmation gates. <br>\n\n### Deployment Geography for Use: <br>\nGlobal <br>\n\n## Known Risks and Mitigations: <br>\nRisk: DeFi workflow plans can lead to financial loss if stale quotes, market movement, or incomplete companion-skill coverage are ignored. <br>\nMitigation: Verify each quote and required companion skill before confirmation, and keep execution in separately reviewed skills with their own confirmation gates. <br>\nRisk: Session workflow records and optional tokens may expose paid or session-tracking features when configured. <br>\nMitigation: Provide FARMDASH_API_KEY or session tokens only when intentionally using those features, and do not provide wallet secrets or signing material. <br>\n\n\n## Reference(s): <br>\n- [ClawHub release page](https://clawhub.ai/parmasanandgarlic/skills/farmdash-trail-marshal) <br>\n- [FarmDash agents homepage](https://www.farmdash.one/agents) <br>\n- [FarmDash API schema](https://www.farmdash.one/agents/openapi.yaml) <br>\n- [FarmDash MCP server config](https://www.farmdash.one/.well-known/mcp.json) <br>\n\n\n## Skill Output: <br>\n**Output Type(s):** [text, markdown, configuration, guidance] <br>\n**Output Format:** [Markdown guidance with structured workflow metadata and status records] <br>\n**Output Parameters:** [1D] <br>\n**Other Properties Related to Output:** [Plans are advisory orchestration outputs; state-changing execution remains delegated to separately installed skills.] <br>\n\n## Skill Version(s): <br>\n0.1.7 (source: server release metadata) <br>\n\n## Ethical Considerations: <br>\nUsers should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment. <br>\n\nFile v0.1.7:SKILL.json\n\n{\n  \"name\": \"FarmDash Trail Marshal\",\n  \"version\": \"1.0.1\",\n  \"description\": \"Guarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action — every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate.\",\n  \"tags\": [\n    \"defi\",\n    \"ai-agent\",\n    \"autonomous-agent\",\n    \"openclaw\",\n    \"clawhub\",\n    \"mcp\",\n    \"crypto\",\n    \"web3\",\n    \"onchain\",\n    \"orchestration\",\n    \"workflow\",\n    \"agent-workflows\",\n    \"multi-agent\",\n    \"quality-gates\",\n    \"risk-management\",\n    \"yield-farming\",\n    \"airdrop\",\n    \"read-only\",\n    \"zero-custody\",\n    \"farmdash\",\n    \"trail-marshal\"\n  ],\n  \"author\": \"FarmDash Pioneers (@Parmasanandgarlic)\",\n  \"homepage\": \"https://www.farmdash.one/agents\",\n  \"icon\": \"🪪\",\n  \"tools\": [\n    {\n      \"name\": \"list_workflows\",\n      \"description\": \"Returns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\\n\\n**\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"plan_workflow\",\n      \"description\": \"Builds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\\n\\n- available steps.\\n- missing steps.\\n- state-changing steps.\\n- confirmation count.\\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\\n\\nUse this before telling the user a workflow is executable.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"run_workflow\",\n      \"description\": \"Creates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\\n\\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"get_workflow_status\",\n      \"description\": \"Reads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    }\n  ],\n  \"operatingRules\": [],\n  \"requiredEnv\": {\n    \"FARMDASH_API_KEY\": {\n      \"description\": \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\",\n      \"required\": false\n    }\n  }\n}\n\nArchive v0.1.6: 4 files, 9446 bytes\n\nFiles: skill-card.md (3131b), SKILL.json (3230b), SKILL.md (15457b), _meta.json (141b)\n\nFile v0.1.6:SKILL.md\n\n---\nname: FarmDash Trail Marshal\ndescription: \"Guarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action — every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate.\"\ntags: [\"defi\", \"ai-agent\", \"autonomous-agent\", \"openclaw\", \"clawhub\", \"mcp\", \"crypto\", \"web3\", \"onchain\", \"orchestration\", \"workflow\", \"agent-workflows\", \"multi-agent\", \"quality-gates\", \"risk-management\", \"yield-farming\", \"airdrop\", \"read-only\", \"zero-custody\", \"farmdash\", \"trail-marshal\"]\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\nhomepage: https://www.farmdash.one\nversion: \"1.0.1\"\nicon: 🪪\nenv:\n  FARMDASH_API_KEY:\n    description: \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\"\n    required: false\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"primaryEnv\":\"FARMDASH_API_KEY\",\"apiKeyRequired\":false,\"freeScoutKeyless\":true,\"freeScoutKey\":\"fd_scout_free\",\"execution\":\"guarded-orchestration-no-execution\"}}\n---\n\n# FarmDash Trail Marshal\n\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\n\n## What this skill does\n\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\n\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\n\n## Available tools\n\n### `list_workflows`\n\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\n\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\n\n**Returns shape:**\n\n```\n{\n  \"workflows\": [\n    {\n      \"id\": \"farm_hyperliquid\",\n      \"name\": \"Hyperliquid Point Farming\",\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\n      \"requiredTier\": \"syndicate\",\n      \"steps\": [\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\n      ]\n    }\n  ]\n}\n```\n\nThe detailed `steps` for each recipe are delivered at runtime as part of the `list_workflows` JSON response. Agents fetch the catalog once per session and cache it for no longer than one hour, or five minutes when a user is actively approving execution.\n\nCodebase alignment note: Wagon steps use `get_wallet_balances`; do not use the stale alias `get_balances`.\n\n**Agent posture:** Call `list_workflows` once at conversation start. Present recipes by name when the user describes a goal that matches one of them.\n\n### `plan_workflow`\n\nBuilds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\n\n- available steps.\n- missing steps.\n- state-changing steps.\n- confirmation count.\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\n\nUse this before telling the user a workflow is executable.\n\n### `run_workflow`\n\nCreates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\n\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\n\n### `get_workflow_status`\n\nReads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\n\n## Workflow catalog (high level)\n\nEighteen recipes are published. Each is **orchestration metadata** — Trail Marshal never executes any state-changing step itself; the user's separately-installed sub-skills do, under their own ClawScan-reviewed contracts.\n\n| ID | Goal | Tier | User confirmations |\n|---|---|---|---|\n| `farm_hyperliquid` | Coordinate Hyperliquid points, spot exposure, and perps/funding research | Syndicate | per execution step |\n| `rotate_quarterly` | Compare current positions against fresh high-heat opportunities | Pioneer | per swap |\n| `protect_portfolio` | Risk-first scan and optional emergency mitigation | Pioneer | 0 if no action |\n| `tax_loss_harvest` | Identify loss-harvest candidates and quote realization paths | Pioneer | per swap |\n| `delta_neutral_setup` | Coordinate paired spot + perp exposure | Syndicate | spot leg + perps leg |\n| `idle_capital_deploy` | Find idle stables/native assets and compare deployment targets | Scout catalog / Pioneer wallet data | per swap |\n| `sybil_dilution` | Review sybil-risk clusters and propose lower-correlation behavior | Pioneer | per action |\n| `yield_optimization` | Review harvest/redeposit candidates against Trail Heat | Pioneer | per swap |\n| `emergency_exit` | Plan a protocol exit and unwind route | Scout catalog / Pioneer wallet data | per swap |\n| `rebalance_portfolio` | Re-align portfolio to target weights and current Trail Heat | Pioneer | per swap |\n| `stablecoin_yield_ladder` | Build a risk-tiered stablecoin deployment ladder | Pioneer | per deposit intent |\n| `post_trade_ledger_review` | Reconcile activity and export records after execution | Pioneer | 0 |\n| `pre_trade_edge_audit` | Confirm positive net edge, clean route quality, fresh risk checks, and quote stability before execution | Pioneer | 0 before execution handoff |\n| `post_execution_quality_review` | Compare expected versus realized output before allowing dependent follow-on actions | Pioneer | 0 |\n| `perps_liquidation_buffer_check` | Reject or resize perps orders that do not survive normal volatility around liquidation | Pioneer | 0 before execution handoff |\n| `funding_carry_break_even_audit` | Validate funding trades after fees, slippage, margin pressure, and funding flip risk | Pioneer | 0 before execution handoff |\n| `session_command_center` | Coordinate context, event snapshots, workflow plans, and heartbeats | Pioneer | per execution handoff |\n| `hedged_airdrop_rotation` | Rotate into high-heat farms while planning hedge coverage | Syndicate | spot leg + hedge leg |\n\nEach recipe's full step graph is returned at runtime by `list_workflows`. The runtime sequence references the user's own separately-installed sub-skills (e.g. *FarmDash Signal Architect* for spot routing, *FarmDash Futures Strategist* for perps); Trail Marshal does not bundle, invoke, or import them.\n\n## Permissions\n\nTrail Marshal can read the public workflow catalog without session state. `run_workflow` requires the user's FarmDash agent `sessionToken` because it creates a session-scoped workflow run record. It cannot modify wallet state, approve allowances, or initiate any on-chain action.\n\nIf a recipe in the catalog requires execution, the user's separately-installed sub-skill performs that work under its own ClawScan-reviewed contract — Trail Marshal stays out of the execution call path.\n\nIf the runtime catalog references a skill that is not installed in the user's environment, such as `farmdash-camp-guard`, `farmdash-supply-master`, `farmdash-hedge-warden`, or `farmdash-ledger-keeper`, mark that step as unavailable and continue with the safe subset. Never fabricate a missing tool call.\n\n## Tier model\n\n| Tier | Cost | Limits | Capability |\n|---|---|---|---|\n| **Scout** | Free | 5 req / 24h | Returns the public workflow catalog |\n| **Pioneer** | $39.99/mo | 500 req / day | Adds higher-rate catalog use; filtering is via the `filter` search string |\n| **Syndicate** | $199/mo | 50k req / day | Higher-rate workflow orchestration and perps-heavy recipes |\n\n## Composition patterns\n\n### Workflow Quality Gate (v0.2)\n\nBefore presenting a recipe as executable, classify every step:\n\n```json\n{\n  \"workflowId\": \"rotate_quarterly\",\n  \"availableSteps\": [],\n  \"missingSteps\": [],\n  \"stateChangingSteps\": [],\n  \"confirmationsRequired\": 0,\n  \"safeSubset\": true,\n  \"fallback\": \"research_only | analysis_only | halt\"\n}\n```\n\nRules:\n\n- A workflow is executable only when every state-changing step has its owning execution skill installed.\n- A missing read-only step lowers confidence; a missing execution step changes the recipe to `analysis_only`.\n- If the recipe includes `execute_swap` or `execute_perp_order`, quote the confirmation count as \"per execution step\" unless the live catalog supplies a stricter count.\n- If more than five minutes pass after the user reviews a plan, re-fetch `list_workflows` and re-run the read-only sense steps.\n\n### The Sense / Decide / Present / Verify loop\n\nEvery Trail Marshal recipe respects this 4-phase pattern:\n\n```\nSENSE    → Read-only research and portfolio skills gather state\nDECIDE   → The recipe ranks options and presents a plan to the user\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle\n```\n\nIf a recipe skips SENSE, it is unsafe. If it skips VERIFY, the agent never improves. Trail Marshal recipes always include both.\n\n### Confirmation gate\n\nTrail Marshal cannot pre-confirm anything. Each user confirmation in a recipe happens at the moment the user's *separate* sub-skill presents its quote — not in Trail Marshal's call path. If asked to run a recipe without intermediate review, the agent should refuse and explain the contract.\n\n### Tier composition\n\nA recipe's required tier is the **maximum** of its sub-skill tiers. When a user is one tier below a recipe's requirement, the agent should recognize the gap from `requiredTier`, offer the lower-tier subset of the recipe (e.g. research-only without the execution leg), and surface an upgrade link if the user wants the full version. Never silently downgrade a recipe without telling the user what is being skipped.\n\n## Reasoning guidelines for agents\n\n- **Speak in workflows, not tool soup.** Quote a recipe's name and goal when presenting a plan.\n- **Quote the `confirms` count up front** so the user knows how many user-confirmation steps the recipe involves.\n- **Re-fetch and re-present** if more than ~5 minutes pass between catalog read and the user's review.\n- **No new analysis.** Trail Marshal returns recipe metadata; it does not editorialize or override what other skills produce.\n\n## Risk warnings the agent should surface\n\nFor any recipe whose execution legs are owned by a separately-installed sub-skill, the agent should restate to the user that:\n\n- DeFi positions can lose value rapidly; airdrop rewards are speculative and not guaranteed.\n- Smart contracts can be exploited; even high-Trail-Heat protocols carry technical risk.\n- Cross-chain transfers are irreversible if the wrong address or chain is selected.\n- Slippage, gas, MEV, and routing fees materially affect realized returns.\n- Past Trail Heat performance does not guarantee future scores.\n- The user retains full responsibility for every on-chain decision they confirm via their own execution skill.\n\n## Commercial disclosure (inherited from sub-skills)\n\nWhen a recipe surfaces a `https://www.farmdash.one/go/{slug}` partnership route, the standard FarmDash commercial disclosure must be included in the same message: FarmDash may receive referral, affiliate, or routing compensation, and fee details live at `https://www.farmdash.one/fees`. Trail Marshal does not introduce new partner routes of its own.\n\n## Versioning\n\n**v1.0 (current).** Catalog plus guarded orchestrator tools: `plan_workflow`, `run_workflow`, and `get_workflow_status`. The migration is strictly additive; `list_workflows` remains unchanged for existing agents.\n\n## Disclaimers\n\nWorkflow recipes are documentation, not financial advice. Any action performed by separately-installed sub-skills is the user's decision under that sub-skill's own contract. Trail Marshal performs no on-chain action; the user retains full responsibility for every decision they confirm.\n\n---\n\n**Install:** Copy this file into your OpenClaw workspace, or fetch `https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md`.\n\n**Companion skills (separately installed by the user):** FarmDash Trail Intelligence (research), FarmDash Wagon Steward (read-only portfolio), FarmDash Signal Architect (spot routing), FarmDash Futures Strategist (perps), FarmDash Camp Guard (risk), FarmDash Supply Master (yield), FarmDash Hedge Warden (hedges), FarmDash Ledger Keeper (records), FarmDash Autonomous Operator (session context).\n\n**Dashboard:** `https://www.farmdash.one`\n**Agent Hub:** `https://www.farmdash.one/agents`\n**MCP Config:** `https://www.farmdash.one/.well-known/mcp.json`\n\n---\n\n## Optional Setup Check\n\nNo registration call is required to install this skill, browse workflow recipes, or use research-only planning. Trail Marshal's core orchestration docs remain usable without onboarding and without `FARMDASH_API_KEY`.\n\nOnly run the setup check if the user explicitly asks to verify FarmDash tier/setup status and agrees to send the listed metadata.\n\nData sent:\n\n- public `agentAddress`\n- `X-ClawHub-Skill: farmdash-trail-marshal`\n\nData not sent:\n\n- private keys, seed phrases, mnemonics, wallet exports, OAuth tokens, raw wallet secrets, or signed transactions\n\nOptional command after consent:\n\n```bash\ncurl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\n  -H \"Content-Type: application/json\" \\\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'\n```\n\nThis returns tier status and available tool access. Skipping this step does not disable the workflow catalog or research-only planning.\n\n### Next steps:\n1. Use the Scout workflow catalog without setup when no key is configured.\n2. Add `FARMDASH_API_KEY` only when the user wants higher-rate Pioneer/Syndicate workflow features.\n3. Browse the OpenAPI spec at `https://www.farmdash.one/agents/openapi.yaml`.\n\nFile v0.1.6:_meta.json\n\n{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.6\",\n  \"publishedAt\": 1780185143604\n}\n\nFile v0.1.6:skill-card.md\n\n## Description: <br>\nGuarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action; every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate. <br>\n\nThis skill is ready for commercial/non-commercial use. <br>\n\n## Publisher: <br>\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic) <br>\n\n### License/Terms of Use: <br>\nMIT-0 <br>\n\n\n## Use Case: <br>\nExternal OpenClaw agents use this skill to list FarmDash DeFi workflow recipes, check which companion skills are available, plan quality gates, and record guarded session workflow status before any separate execution skill asks the user for approval. <br>\n\n### Deployment Geography for Use: <br>\nGlobal <br>\n\n## Known Risks and Mitigations: <br>\nRisk: The skill operates in a DeFi planning context and may use FarmDash session tokens and public wallet identifiers for run records. <br>\nMitigation: Use it as a workflow planning layer only, keep private keys, seed phrases, wallet exports, and signed transactions out of the skill, and review companion execution skills separately. <br>\nRisk: A planned workflow may include trading, swapping, bridging, deposits, or perps actions handled by separately installed companion skills. <br>\nMitigation: Require the owning execution skill to present fresh data and obtain explicit user confirmation for each state-changing step before any action is taken. <br>\nRisk: Missing companion skills can make a workflow only partially available or analysis-only. <br>\nMitigation: Classify available, missing, and state-changing steps before presenting a workflow as executable, and continue only with the safe subset when required skills are absent. <br>\n\n\n## Reference(s): <br>\n- [ClawHub Skill Page](https://clawhub.ai/parmasanandgarlic/farmdash-trail-marshal) <br>\n- [Publisher Profile](https://clawhub.ai/user/parmasanandgarlic) <br>\n- [FarmDash Agent Hub](https://www.farmdash.one/agents) <br>\n- [FarmDash MCP Configuration](https://www.farmdash.one/.well-known/mcp.json) <br>\n- [FarmDash OpenAPI Specification](https://www.farmdash.one/agents/openapi.yaml) <br>\n\n\n## Skill Output: <br>\n**Output Type(s):** [guidance, markdown, configuration, shell commands] <br>\n**Output Format:** [Markdown guidance with JSON examples and optional shell commands] <br>\n**Output Parameters:** [1D] <br>\n**Other Properties Related to Output:** [Returns workflow catalog, planning classifications, guarded run records, and workflow status guidance; it does not produce signed transactions or execute on-chain actions.] <br>\n\n## Skill Version(s): <br>\n0.1.6 (source: server-resolved release metadata) <br>\n\n## Ethical Considerations: <br>\nUsers should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment. <br>\n\nFile v0.1.6:SKILL.json\n\n{\n  \"name\": \"FarmDash Trail Marshal\",\n  \"version\": \"1.0.1\",\n  \"description\": \"Guarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action — every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate.\",\n  \"tags\": [\n    \"defi\",\n    \"ai-agent\",\n    \"autonomous-agent\",\n    \"openclaw\",\n    \"clawhub\",\n    \"mcp\",\n    \"crypto\",\n    \"web3\",\n    \"onchain\",\n    \"orchestration\",\n    \"workflow\",\n    \"agent-workflows\",\n    \"multi-agent\",\n    \"quality-gates\",\n    \"risk-management\",\n    \"yield-farming\",\n    \"airdrop\",\n    \"read-only\",\n    \"zero-custody\",\n    \"farmdash\",\n    \"trail-marshal\"\n  ],\n  \"author\": \"FarmDash Pioneers (@Parmasanandgarlic)\",\n  \"homepage\": \"https://www.farmdash.one\",\n  \"icon\": \"🪪\",\n  \"tools\": [\n    {\n      \"name\": \"list_workflows\",\n      \"description\": \"Returns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\\n\\n**\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"plan_workflow\",\n      \"description\": \"Builds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\\n\\n- available steps.\\n- missing steps.\\n- state-changing steps.\\n- confirmation count.\\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\\n\\nUse this before telling the user a workflow is executable.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"run_workflow\",\n      \"description\": \"Creates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\\n\\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"get_workflow_status\",\n      \"description\": \"Reads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {}\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    }\n  ],\n  \"operatingRules\": [],\n  \"requiredEnv\": {\n    \"FARMDASH_API_KEY\": {\n      \"description\": \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\",\n      \"required\": false\n    }\n  }\n}\n\nArchive v0.1.5: 3 files, 7888 bytes\n\nFiles: skill-card.md (2975b), SKILL.md (15129b), _meta.json (141b)\n\nFile v0.1.5:SKILL.md\n\n---\nname: FarmDash Trail Marshal\ndescription: \"Guarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action — every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate.\"\ntags: [\"defi\", \"ai-agent\", \"autonomous-agent\", \"openclaw\", \"clawhub\", \"mcp\", \"crypto\", \"web3\", \"onchain\", \"orchestration\", \"workflow\", \"agent-workflows\", \"multi-agent\", \"quality-gates\", \"risk-management\", \"yield-farming\", \"airdrop\", \"read-only\", \"zero-custody\", \"farmdash\", \"trail-marshal\"]\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\nhomepage: https://www.farmdash.one\nversion: \"1.0.0\"\nicon: 🪪\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"execution\":\"guarded-orchestration-no-execution\"}}\n---\n\n# FarmDash Trail Marshal\n\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\n\n## What this skill does\n\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\n\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\n\n## Available tools\n\n### `list_workflows`\n\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\n\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\n\n**Returns shape:**\n\n```\n{\n  \"workflows\": [\n    {\n      \"id\": \"farm_hyperliquid\",\n      \"name\": \"Hyperliquid Point Farming\",\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\n      \"requiredTier\": \"syndicate\",\n      \"steps\": [\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\n      ]\n    }\n  ]\n}\n```\n\nThe detailed `steps` for each recipe are delivered at runtime as part of the `list_workflows` JSON response. Agents fetch the catalog once per session and cache it for no longer than one hour, or five minutes when a user is actively approving execution.\n\nCodebase alignment note: Wagon steps use `get_wallet_balances`; do not use the stale alias `get_balances`.\n\n**Agent posture:** Call `list_workflows` once at conversation start. Present recipes by name when the user describes a goal that matches one of them.\n\n### `plan_workflow`\n\nBuilds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\n\n- available steps.\n- missing steps.\n- state-changing steps.\n- confirmation count.\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\n\nUse this before telling the user a workflow is executable.\n\n### `run_workflow`\n\nCreates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\n\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\n\n### `get_workflow_status`\n\nReads a workflow run by `runId`. Use this after a long-running loop, after a context restore, or before resuming a paused autonomous session.\n\n## Workflow catalog (high level)\n\nEighteen recipes are published. Each is **orchestration metadata** — Trail Marshal never executes any state-changing step itself; the user's separately-installed sub-skills do, under their own ClawScan-reviewed contracts.\n\n| ID | Goal | Tier | User confirmations |\n|---|---|---|---|\n| `farm_hyperliquid` | Coordinate Hyperliquid points, spot exposure, and perps/funding research | Syndicate | per execution step |\n| `rotate_quarterly` | Compare current positions against fresh high-heat opportunities | Pioneer | per swap |\n| `protect_portfolio` | Risk-first scan and optional emergency mitigation | Pioneer | 0 if no action |\n| `tax_loss_harvest` | Identify loss-harvest candidates and quote realization paths | Pioneer | per swap |\n| `delta_neutral_setup` | Coordinate paired spot + perp exposure | Syndicate | spot leg + perps leg |\n| `idle_capital_deploy` | Find idle stables/native assets and compare deployment targets | Scout catalog / Pioneer wallet data | per swap |\n| `sybil_dilution` | Review sybil-risk clusters and propose lower-correlation behavior | Pioneer | per action |\n| `yield_optimization` | Review harvest/redeposit candidates against Trail Heat | Pioneer | per swap |\n| `emergency_exit` | Plan a protocol exit and unwind route | Scout catalog / Pioneer wallet data | per swap |\n| `rebalance_portfolio` | Re-align portfolio to target weights and current Trail Heat | Pioneer | per swap |\n| `stablecoin_yield_ladder` | Build a risk-tiered stablecoin deployment ladder | Pioneer | per deposit intent |\n| `post_trade_ledger_review` | Reconcile activity and export records after execution | Pioneer | 0 |\n| `pre_trade_edge_audit` | Confirm positive net edge, clean route quality, fresh risk checks, and quote stability before execution | Pioneer | 0 before execution handoff |\n| `post_execution_quality_review` | Compare expected versus realized output before allowing dependent follow-on actions | Pioneer | 0 |\n| `perps_liquidation_buffer_check` | Reject or resize perps orders that do not survive normal volatility around liquidation | Pioneer | 0 before execution handoff |\n| `funding_carry_break_even_audit` | Validate funding trades after fees, slippage, margin pressure, and funding flip risk | Pioneer | 0 before execution handoff |\n| `session_command_center` | Coordinate context, event snapshots, workflow plans, and heartbeats | Pioneer | per execution handoff |\n| `hedged_airdrop_rotation` | Rotate into high-heat farms while planning hedge coverage | Syndicate | spot leg + hedge leg |\n\nEach recipe's full step graph is returned at runtime by `list_workflows`. The runtime sequence references the user's own separately-installed sub-skills (e.g. *FarmDash Signal Architect* for spot routing, *FarmDash Futures Strategist* for perps); Trail Marshal does not bundle, invoke, or import them.\n\n## Permissions\n\nTrail Marshal can read the public workflow catalog without session state. `run_workflow` requires the user's FarmDash agent `sessionToken` because it creates a session-scoped workflow run record. It cannot modify wallet state, approve allowances, or initiate any on-chain action.\n\nIf a recipe in the catalog requires execution, the user's separately-installed sub-skill performs that work under its own ClawScan-reviewed contract — Trail Marshal stays out of the execution call path.\n\nIf the runtime catalog references a skill that is not installed in the user's environment, such as `farmdash-camp-guard`, `farmdash-supply-master`, `farmdash-hedge-warden`, or `farmdash-ledger-keeper`, mark that step as unavailable and continue with the safe subset. Never fabricate a missing tool call.\n\n## Tier model\n\n| Tier | Cost | Limits | Capability |\n|---|---|---|---|\n| **Scout** | Free | 5 req / 24h | Returns the public workflow catalog |\n| **Pioneer** | $29/mo | 500 req / day | Adds higher-rate catalog use; filtering is via the `filter` search string |\n| **Syndicate** | $199/mo | 50k req / day | Higher-rate workflow orchestration and perps-heavy recipes |\n\n## Composition patterns\n\n### Workflow Quality Gate (v0.2)\n\nBefore presenting a recipe as executable, classify every step:\n\n```json\n{\n  \"workflowId\": \"rotate_quarterly\",\n  \"availableSteps\": [],\n  \"missingSteps\": [],\n  \"stateChangingSteps\": [],\n  \"confirmationsRequired\": 0,\n  \"safeSubset\": true,\n  \"fallback\": \"research_only | analysis_only | halt\"\n}\n```\n\nRules:\n\n- A workflow is executable only when every state-changing step has its owning execution skill installed.\n- A missing read-only step lowers confidence; a missing execution step changes the recipe to `analysis_only`.\n- If the recipe includes `execute_swap` or `execute_perp_order`, quote the confirmation count as \"per execution step\" unless the live catalog supplies a stricter count.\n- If more than five minutes pass after the user reviews a plan, re-fetch `list_workflows` and re-run the read-only sense steps.\n\n### The Sense / Decide / Present / Verify loop\n\nEvery Trail Marshal recipe respects this 4-phase pattern:\n\n```\nSENSE    → Read-only research and portfolio skills gather state\nDECIDE   → The recipe ranks options and presents a plan to the user\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle\n```\n\nIf a recipe skips SENSE, it is unsafe. If it skips VERIFY, the agent never improves. Trail Marshal recipes always include both.\n\n### Confirmation gate\n\nTrail Marshal cannot pre-confirm anything. Each user confirmation in a recipe happens at the moment the user's *separate* sub-skill presents its quote — not in Trail Marshal's call path. If asked to run a recipe without intermediate review, the agent should refuse and explain the contract.\n\n### Tier composition\n\nA recipe's required tier is the **maximum** of its sub-skill tiers. When a user is one tier below a recipe's requirement, the agent should recognize the gap from `requiredTier`, offer the lower-tier subset of the recipe (e.g. research-only without the execution leg), and surface an upgrade link if the user wants the full version. Never silently downgrade a recipe without telling the user what is being skipped.\n\n## Reasoning guidelines for agents\n\n- **Speak in workflows, not tool soup.** Quote a recipe's name and goal when presenting a plan.\n- **Quote the `confirms` count up front** so the user knows how many user-confirmation steps the recipe involves.\n- **Re-fetch and re-present** if more than ~5 minutes pass between catalog read and the user's review.\n- **No new analysis.** Trail Marshal returns recipe metadata; it does not editorialize or override what other skills produce.\n\n## Risk warnings the agent should surface\n\nFor any recipe whose execution legs are owned by a separately-installed sub-skill, the agent should restate to the user that:\n\n- DeFi positions can lose value rapidly; airdrop rewards are speculative and not guaranteed.\n- Smart contracts can be exploited; even high-Trail-Heat protocols carry technical risk.\n- Cross-chain transfers are irreversible if the wrong address or chain is selected.\n- Slippage, gas, MEV, and routing fees materially affect realized returns.\n- Past Trail Heat performance does not guarantee future scores.\n- The user retains full responsibility for every on-chain decision they confirm via their own execution skill.\n\n## Commercial disclosure (inherited from sub-skills)\n\nWhen a recipe surfaces a `https://www.farmdash.one/go/{slug}` partnership route, the standard FarmDash commercial disclosure must be included in the same message: FarmDash may receive referral, affiliate, or routing compensation, and fee details live at `https://www.farmdash.one/fees`. Trail Marshal does not introduce new partner routes of its own.\n\n## Versioning\n\n**v1.0 (current).** Catalog plus guarded orchestrator tools: `plan_workflow`, `run_workflow`, and `get_workflow_status`. The migration is strictly additive; `list_workflows` remains unchanged for existing agents.\n\n## Disclaimers\n\nWorkflow recipes are documentation, not financial advice. Any action performed by separately-installed sub-skills is the user's decision under that sub-skill's own contract. Trail Marshal performs no on-chain action; the user retains full responsibility for every decision they confirm.\n\n---\n\n**Install:** Copy this file into your OpenClaw workspace, or fetch `https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md`.\n\n**Companion skills (separately installed by the user):** FarmDash Trail Intelligence (research), FarmDash Wagon Steward (read-only portfolio), FarmDash Signal Architect (spot routing), FarmDash Futures Strategist (perps), FarmDash Camp Guard (risk), FarmDash Supply Master (yield), FarmDash Hedge Warden (hedges), FarmDash Ledger Keeper (records), FarmDash Autonomous Operator (session context).\n\n**Dashboard:** `https://www.farmdash.one`\n**Agent Hub:** `https://www.farmdash.one/agents`\n**MCP Config:** `https://www.farmdash.one/.well-known/mcp.json`\n\n---\n\n## Optional Setup Check\n\nNo registration call is required to install this skill, browse workflow recipes, or use research-only planning. Trail Marshal's core orchestration docs remain usable without onboarding and without `FARMDASH_API_KEY`.\n\nOnly run the setup check if the user explicitly asks to verify FarmDash tier/setup status and agrees to send the listed metadata.\n\nData sent:\n\n- public `agentAddress`\n- `X-ClawHub-Skill: farmdash-trail-marshal`\n\nData not sent:\n\n- private keys, seed phrases, mnemonics, wallet exports, OAuth tokens, raw wallet secrets, or signed transactions\n\nOptional command after consent:\n\n```bash\ncurl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\n  -H \"Content-Type: application/json\" \\\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'\n```\n\nThis returns tier status and available tool access. Skipping this step does not disable the workflow catalog or research-only planning.\n\n### Next steps:\n1. Use the Scout workflow catalog without setup when no key is configured.\n2. Add `FARMDASH_API_KEY` only when the user wants higher-rate Pioneer/Syndicate workflow features.\n3. Browse the OpenAPI spec at `https://www.farmdash.one/agents/openapi.yaml`.\n\nFile v0.1.5:_meta.json\n\n{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.5\",\n  \"publishedAt\": 1779662291401\n}\n\nFile v0.1.5:skill-card.md\n\n## Description: <br>\nGuarded DeFi orchestration layer for OpenClaw agents. Lists named multi-skill workflow recipes, builds quality gates, creates session-scoped workflow run records, and reports workflow status. Trail Marshal holds no keys and performs no on-chain action - every state-changing step remains delegated to the user's separately-installed execution skill under that skill's own confirmation gate. <br>\n\nThis skill is ready for commercial/non-commercial use. <br>\n\n## Publisher: <br>\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic) <br>\n\n### License/Terms of Use: <br>\nMIT-0 <br>\n\n\n## Use Case: <br>\nExternal users and OpenClaw agent developers use this skill to plan FarmDash DeFi workflows, check required companion skills, create guarded workflow run records, and report workflow status without directly executing wallet-changing actions. <br>\n\n### Deployment Geography for Use: <br>\nGlobal <br>\n\n## Known Risks and Mitigations: <br>\nRisk: The skill supports DeFi workflow planning and can require wallet or session metadata, which may expose sensitive operational context to FarmDash. <br>\nMitigation: Install only when comfortable sharing the disclosed metadata, and do not provide private keys, seed phrases, wallet exports, OAuth tokens, or raw wallet secrets. <br>\nRisk: Workflow plans may include swaps, deposits, bridges, approvals, or perps actions that affect wallet state when performed by a separate execution skill. <br>\nMitigation: Require a fresh explicit confirmation in the separately installed execution skill before any walle\n\nArchive v0.1.4: 2 files, 6309 bytes\n\nFiles: SKILL.md (14856b), _meta.json (141b)\n\nArchive v0.1.3: 2 files, 5994 bytes\n\nFiles: SKILL.md (13989b), _meta.json (141b)\n\nArchive v0.1.2: 2 files, 5067 bytes\n\nFiles: SKILL.md (11698b), _meta.json (141b)\n\nArchive v0.1.1: 2 files, 4384 bytes\n\nFiles: SKILL.md (9939b), _meta.json (141b)","readmeExcerpt":"Skill: FarmDash Trail Marshal Owner: parmasanandgarlic Summary: Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only. Tags: agent:0.1.1, agent-orchestration:0.1.9, agent-recipes:0.1.6, agent-workflows:0.1.7, agentic defi:0.1.10, ai agent:0.1.10, ai agent orchestration layer:0.1.10, ai agent risk management:0.1.10, ai-ag","codeSnippets":[],"executableExamples":[{"language":"text","snippet":"{\n  \"workflows\": [\n    {\n      \"id\": \"farm_hyperliquid\",\n      \"name\": \"Hyperliquid Point Farming\",\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\n      \"requiredTier\": \"syndicate\",\n      \"steps\": [\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\n      ]\n    }\n  ]\n}"},{"language":"json","snippet":"{\n  \"workflowId\": \"rotate_quarterly\",\n  \"availableSteps\": [],\n  \"missingSteps\": [],\n  \"stateChangingSteps\": [],\n  \"confirmationsRequired\": 0,\n  \"safeSubset\": true,\n  \"fallback\": \"research_only | analysis_only | halt\"\n}"},{"language":"text","snippet":"SENSE    → Read-only research and portfolio skills gather state\nDECIDE   → The recipe ranks options and presents a plan to the user\nPRESENT  → User reviews; the user's separately-installed sub-skill (if any) handles confirmation under its own contract\nVERIFY   → Read-only portfolio skill confirms the new state on the next cycle"},{"language":"bash","snippet":"curl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\n  -H \"Content-Type: application/json\" \\\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'"},{"language":"bash","snippet":"curl -X POST https://www.farmdash.one/api/v1/agent/onboard \\\n  -H \"Content-Type: application/json\" \\\n  -H \"X-ClawHub-Skill: farmdash-trail-marshal\" \\\n  -d '{\"agentAddress\": \"0xYOUR_AGENT_WALLET\"}'"},{"language":"text","snippet":"{\n  \"workflows\": [\n    {\n      \"id\": \"farm_hyperliquid\",\n      \"name\": \"Hyperliquid Point Farming\",\n      \"description\": \"Optimize points on Hyperliquid through perps volume, spot balances, and funding arbitrage.\",\n      \"requiredTier\": \"syndicate\",\n      \"steps\": [\n        { \"skill\": \"trail-intelligence\", \"tool\": \"get_trail_heat\", \"purpose\": \"Check HL momentum\" },\n        { \"skill\": \"wagon-steward\", \"tool\": \"get_wallet_balances\", \"purpose\": \"Check current exposure\" },\n        { \"skill\": \"futures-strategist\", \"tool\": \"scan_funding_rates\", \"purpose\": \"Identify arb opportunities\" }\n      ]\n    }\n  ]\n}"}],"parameters":null,"dependencies":[],"permissions":[],"extractedFiles":[{"path":"SKILL.md","content":"---\nname: FarmDash Trail Marshal\ndescription: \"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\"\ntags: [\"defi\",\"defi-workflows\",\"agent-orchestration\",\"workflow-automation\",\"multi-agent-system\",\"ai-agent-workflows\",\"policy-gated\",\"supervised-automation\",\"defi-playbooks\",\"quality-gates\",\"workflow-status\",\"session-management\",\"risk-management\",\"yield-farming-workflow\",\"airdrop-workflow\",\"zero-custody\",\"openclaw\",\"mcp\",\"web3\",\"farmdash\"]\nauthor: FarmDash Pioneers (@Parmasanandgarlic)\nhomepage: https://www.farmdash.one/agents\nversion: \"1.1.2\"\nicon: 🪪\nenv:\n  FARMDASH_API_KEY:\n    description: \"Optional Bearer token for Pioneer or Syndicate workflow features. Scout workflow catalog works without any key or with the public fd_scout_free token.\"\n    required: false\nmetadata: {\"openclaw\":{\"homepage\":\"https://www.farmdash.one/agents\",\"skillKey\":\"farmdash-trail-marshal\",\"primaryEnv\":\"FARMDASH_API_KEY\",\"apiKeyRequired\":false,\"freeScoutKeyless\":true,\"freeScoutKey\":\"fd_scout_free\",\"execution\":\"guarded-orchestration-no-execution\"}}\n---\n\n# FarmDash Trail Marshal\n\n> Use this skill when running a multi-step DeFi goal: pick a named, confirmation-gated recipe instead of improvising the sequence at runtime.\n\n> **Security Posture.** Guarded orchestration only. Trail Marshal exposes `list_workflows`, `plan_workflow`, `run_workflow`, and `get_workflow_status`. It can build a plan and persist a workflow run record, but it cannot sign, approve, bridge, swap, deposit, or place perp orders. Every state-changing step is owned by a separately-installed sub-skill under that sub-skill's confirmation gate.\n\n## What this skill does\n\nTrail Marshal is a **composition and quality-gate layer** for OpenClaw agents. It returns a catalog of named multi-skill workflow recipes, checks whether the required skills are installed, separates read-only steps from state-changing steps, and records guarded workflow runs for session-based agents.\n\nWhen the user says *\"Hyperliquid is hot, set me up,\"* a single skill cannot fully answer. The agent first reads market state and the user's wallet via read-only research skills, presents a plan, and (only with the user's explicit confirmation) hands off to the user's separately-installed execution skills. Trail Marshal documents that orchestration as a named recipe so the agent does not have to invent the sequence at runtime.\n\n## Available tools\n\n### `list_workflows`\n\nReturns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\n\n**Inputs:** Optional `filter` string. The current endpoint searches workflow id, name, description, required tier, step skill, step tool, and step purpose. Legacy `category` or `tier` filters should be converted into a plain `filter` value before calling.\n\n**Returns shape:**\n\n```\n{\n  \"workflows\": [\n    {\n  "},{"path":"_meta.json","content":"{\n  \"ownerId\": \"kn753be85erp4cey7vbq9aq9sd826p37\",\n  \"slug\": \"farmdash-trail-marshal\",\n  \"version\": \"0.1.10\",\n  \"publishedAt\": 1790466329903\n}"},{"path":"skill-card.md","content":"## Description:\n\nCoordinates named DeFi workflow recipes, checks required skills, and records guarded runs without executing wallet transactions.\n\nThis skill is ready for commercial/non-commercial use.\n\n## Publisher:\n\n[parmasanandgarlic](https://clawhub.ai/user/parmasanandgarlic)\n\n### License/Terms of Use:\n\nMIT-0\n\n## Use Case:\n\nDeFi users and their agents use this skill to choose named workflow recipes, assess missing prerequisites and confirmation requirements, and track supervised runs. Wallet-changing steps require separately installed execution skills and explicit user confirmation.\n\n### Deployment Geography for Use:\n\nGlobal\n\n## Known Risks and Mitigations:\n\nRisk: A workflow may suggest financially risky DeFi actions or rely on stale market data.\n\nMitigation: Refresh research and review costs and risks before any handoff; require separate user confirmation for every wallet-changing action.\n\nRisk: Sharing wallet secrets or credentials with an untrusted endpoint could compromise assets or account access.\n\nMitigation: Never provide private keys, seed phrases, signed transactions, or raw wallet secrets; use optional API keys and session tokens only with trusted FarmDash endpoints.\n\n## Reference(s):\n\n- [FarmDash Agent Hub](https://www.farmdash.one/agents)\n- [FarmDash agent integration documentation](https://www.farmdash.one/docs)\n- [FarmDash OpenAPI contract](https://www.farmdash.one/agents/openapi.yaml)\n- [FarmDash MCP discovery manifest](https://www.farmdash.one/.well-known/mcp.json)\n- [FarmDash Trail Marshal skill manual](https://www.farmdash.one/openclaw-skills/farmdash-trail-marshal/SKILL.md)\n\n## Skill Output:\n\n**Output Type(s):** [Guidance, Workflow plans, Workflow status]\n\n**Output Format:** [Text and structured JSON]\n\n**Output Parameters:** [1D]\n\n**Other Properties Related to Output:** [Catalog entries, prerequisite and confirmation checks, and guarded run records; no wallet execution.]\n\n## Skill Version(s):\n\n0.1.10 (source: server-resolved ClawHub release)\n\n## Ethical Considerations:\n\nUsers should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment."},{"path":"SKILL.json","content":"{\n  \"name\": \"FarmDash Trail Marshal\",\n  \"version\": \"1.1.2\",\n  \"description\": \"Use when composing multi-step DeFi workflows from FarmDash tools: plan, run, inspect supervised pipelines. Wallet steps stay policy-gated, preview-only.\",\n  \"tags\": [\n    \"defi\",\n    \"defi-workflows\",\n    \"agent-orchestration\",\n    \"workflow-automation\",\n    \"multi-agent-system\",\n    \"ai-agent-workflows\",\n    \"policy-gated\",\n    \"supervised-automation\",\n    \"defi-playbooks\",\n    \"quality-gates\",\n    \"workflow-status\",\n    \"session-management\",\n    \"risk-management\",\n    \"yield-farming-workflow\",\n    \"airdrop-workflow\",\n    \"zero-custody\",\n    \"openclaw\",\n    \"mcp\",\n    \"web3\",\n    \"farmdash\"\n  ],\n  \"author\": \"FarmDash Pioneers (@Parmasanandgarlic)\",\n  \"homepage\": \"https://www.farmdash.one/agents\",\n  \"icon\": \"🪪\",\n  \"tools\": [\n    {\n      \"name\": \"list_workflows\",\n      \"description\": \"Returns the canonical workflow catalog. Each entry includes a name, goal, required tier, the number of explicit user confirmations the recipe requires, and the sub-skills it composes.\\r\\n\\r\\n**\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"filter\": {\n            \"type\": \"string\"\n          }\n        },\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"plan_workflow\",\n      \"description\": \"Builds the Workflow Quality Gate for a named recipe. Inputs are `workflowId`, optional `installedSkills`, and optional `agentAddress`. The output classifies:\\r\\n\\r\\n- available steps.\\r\\n- missing steps.\\r\\n- state-changing steps.\\r\\n- confirmation count.\\r\\n- fallback mode: `ready`, `research_only`, or `analysis_only`.\\r\\n\\r\\nUse this before telling the user a workflow is executable.\",\n      \"inputSchema\": {\n        \"type\": \"object\",\n        \"properties\": {\n          \"workflowId\": {\n            \"type\": \"string\"\n          },\n          \"installedSkills\": {\n            \"type\": \"array\",\n            \"items\": {\n              \"type\": \"string\"\n            }\n          },\n          \"agentAddress\": {\n            \"type\": \"string\",\n            \"pattern\": \"^0x[a-fA-F0-9]{40}$\"\n          }\n        },\n        \"required\": [\n          \"workflowId\"\n        ],\n        \"additionalProperties\": false,\n        \"$schema\": \"http://json-schema.org/draft-07/schema#\"\n      },\n      \"outputHints\": [],\n      \"category\": \"read\"\n    },\n    {\n      \"name\": \"run_workflow\",\n      \"description\": \"Creates a guarded workflow run record for a live session. Inputs are `workflowId`, `sessionId`, `agentAddress`, `sessionToken`, and optional `installedSkills`.\\r\\n\\r\\nImportant: `run_workflow` does not run execution tools. If a workflow includes `execute_swap`, `resolve_defi_intent`, or `execute_perp_order`, the run status becomes `awaiting_confirmation` and the owning execution skill must still present fresh data and obtain explicit user approval.\",\n      \"inputSchema\": {\n  "}],"languages":[],"docsSourceLabel":"CLAWHUB","editorialOverview":null,"editorialQuality":{"score":100,"threshold":65,"status":"thin","wordCount":1675,"uniquenessScore":41,"reasons":["uniqueness-below-45"]}},"media":{"evidence":{"source":"no-media","verified":false,"confidence":"low","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":"No screenshots, media assets, or demo links are available."},"primaryImageUrl":null,"mediaAssetCount":0,"assets":[],"demoUrl":null},"ownerResources":{"evidence":{"source":"unclaimed","verified":false,"confidence":"low","updatedAt":"2026-10-10T06:10:56.163Z","emptyReason":"This page has not been claimed by the agent owner."},"hasCustomPage":false,"customPageUpdatedAt":null,"customLinks":[],"structuredLinks":{"docsUrl":null,"demoUrl":null,"supportUrl":null,"pricingUrl":null,"statusUrl":null},"customPage":null},"relatedAgents":{"evidence":{"source":"protocol-neighbors","verified":false,"confidence":"medium","updatedAt":"2026-10-10T10:49:59.073Z","emptyReason":null},"items":[{"id":"8ebccd8e-3863-4187-8355-c3f14e1f9edf","entityType":"agent","canonicalPath":"/agent/iofficeai-aionui","slug":"iofficeai-aionui","name":"AionUi","description":"Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!","url":"https://github.com/iOfficeAI/AionUi","homepage":"https://www.aionui.com","source":"GITHUB_REPOS","protocols":["MCP","OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-10-09T19:11:12.944Z","createdAt":"2026-02-25T03:38:16.584Z","downloads":null},{"id":"b917f68a-ebff-438e-84f8-3f4b2494c0bc","entityType":"agent","canonicalPath":"/agent/activepieces-activepieces","slug":"activepieces-activepieces","name":"activepieces","description":"AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents","url":"https://github.com/activepieces/activepieces","homepage":"https://www.activepieces.com","source":"GITHUB_REPOS","protocols":["OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-04-15T02:22:12.426Z","createdAt":"2026-02-25T03:38:12.412Z","downloads":null},{"id":"5cb26759-3a39-483f-94cf-276a98c13bb8","entityType":"agent","canonicalPath":"/agent/cherryhq-cherry-studio","slug":"cherryhq-cherry-studio","name":"cherry-studio","description":"AI productivity studio with smart chat, autonomous agents, and 300+ assistants. Unified access to frontier LLMs","url":"https://github.com/CherryHQ/cherry-studio","homepage":"https://cherry-ai.com","source":"GITHUB_REPOS","protocols":["MCP","OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-04-11T14:38:40.986Z","createdAt":"2026-02-25T03:38:19.379Z","downloads":null},{"id":"6f6582d0-5d76-4f0f-b81d-86520247950b","entityType":"agent","canonicalPath":"/agent/copilotkit-copilotkit","slug":"copilotkit-copilotkit","name":"CopilotKit","description":"The Frontend for Agents & Generative UI. React + Angular","url":"https://github.com/CopilotKit/CopilotKit","homepage":"https://docs.copilotkit.ai","source":"GITHUB_REPOS","protocols":["OPENCLAW"],"capabilities":[],"safetyScore":100,"overallRank":70,"updatedAt":"2026-03-25T09:50:57.846Z","createdAt":"2026-02-25T03:39:14.617Z","downloads":null}],"links":{"hub":"/agent","source":"/agent/source/clawhub","protocols":[{"label":"OpenClaw","href":"/agent/protocol/openclew"}]}}}