{"id":"fde68255-620b-48ec-adff-660655f15e6a","entityType":"agent","slug":"clawhub-xucheng-swarm-layer","name":"Swarm Layer","canonicalUrl":"https://www.xpersona.co/agent/clawhub-xucheng-swarm-layer","canonicalPath":"/agent/clawhub-xucheng-swarm-layer","generatedAt":"2026-10-09T23:52:04.804Z","source":"CLAWHUB","claimStatus":"UNCLAIMED","verificationTier":"NONE","summary":{"evidence":{"source":"editorial-content","verified":true,"confidence":"high","updatedAt":"2026-10-09T20:25:48.962Z","emptyReason":null},"description":"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,... Skill: Swarm Layer Owner: xucheng Summary: OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,... Tags: harness:0.5.12, latest:0.5.12, openclaw:0.5.12, orchestration:0.5.12, swarm:0.5.12, workflow:0.5.12 Version history: v0.5.12 | 2026-06-11T14:12:42.520Z | user Document artifact acceptance (expectedArtifacts) and","descriptionLabel":"Technical summary","evidenceSummary":"Capability contract not published. No trust telemetry is available yet. 2K downloads reported by the source. Last updated 10/9/2026.","installCommand":"clawhub skill install s17cw3n1h91pjjfryne8gypnr583gys6:swarm-layer","sourceUrl":"https://clawhub.ai/xucheng/swarm-layer","homepage":"https://clawhub.ai/xucheng/skills/swarm-layer","primaryLinks":[{"label":"View on ClawHub","url":"https://clawhub.ai/xucheng/swarm-layer","kind":"source"},{"label":"Homepage","url":"https://clawhub.ai/xucheng/skills/swarm-layer","kind":"homepage"}],"safetyScore":84,"overallRank":62,"popularityScore":48,"trustScore":null,"claimedByName":null,"isOwner":false,"seoDescription":"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,... "},"coverage":{"evidence":{"source":"public-profile","verified":false,"confidence":"medium","updatedAt":"2026-10-09T20:25:48.962Z","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":"2026-10-09T20:25:48.962Z","emptyReason":null},"stars":null,"forks":null,"downloads":2028,"packageName":null,"latestVersion":"0.5.12","tractionLabel":"2K downloads"},"release":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-09T20:25:48.962Z","emptyReason":null},"lastUpdatedAt":"2026-10-09T20:25:48.962Z","lastCrawledAt":"2026-10-09T20:25:48.962Z","lastIndexedAt":null,"nextCrawlAt":"2026-10-10T20:25:48.962Z","lastVerifiedAt":null,"highlights":[{"version":"0.5.12","createdAt":"2026-06-11T14:12:42.520Z","changelog":"Document artifact acceptance (expectedArtifacts) and the 0.5.12 conservative ACP failure-mapping baseline.","fileCount":3,"zipByteSize":8088},{"version":"0.5.11","createdAt":"2026-06-05T05:38:54.767Z","changelog":"Republish the Swarm Layer skill with explicit workdir and current 0.5.11 SKILL.md content.","fileCount":3,"zipByteSize":8069},{"version":"0.5.10","createdAt":"2026-06-05T05:20:11.389Z","changelog":"Refresh the Swarm Layer skill for ACP runtime-mode diagnostics and current 0.5.10 release metadata.","fileCount":3,"zipByteSize":7857},{"version":"0.5.9","createdAt":"2026-06-05T03:05:41.271Z","changelog":"Refresh the Swarm Layer skill for npm-project acpx backend discovery and current 0.5.9 release metadata.","fileCount":3,"zipByteSize":4721},{"version":"0.5.8","createdAt":"2026-06-04T07:29:37.866Z","changelog":"Skill-only correction after the 0.5.7 release: publish the current Daily Papers reliability docs with sync-aware run/status commands and remove stale subagent-era wording.","fileCount":3,"zipByteSize":7916},{"version":"0.5.7","createdAt":"2026-06-04T07:26:53.830Z","changelog":"Refresh the Swarm Layer skill for Daily Papers-style cron reliability, sync-aware run/status commands, and current release metadata.","fileCount":3,"zipByteSize":4777},{"version":"0.5.6","createdAt":"2026-05-25T13:56:51.049Z","changelog":"Refresh the Swarm Layer skill for the ClawHub package artifact hotfix and current release metadata.","fileCount":3,"zipByteSize":7683},{"version":"0.5.5","createdAt":"2026-05-25T13:48:54.635Z","changelog":"Refresh the Swarm Layer skill for the npm 11 install-smoke hotfix and current release metadata.","fileCount":2,"zipByteSize":6411}]},"execution":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No published capability contract is available yet."},"installCommand":"clawhub skill install s17cw3n1h91pjjfryne8gypnr583gys6:swarm-layer","setupComplexity":"low","setupSteps":["Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.","Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data."],"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-xucheng-swarm-layer/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/trust"},"curlExamples":["curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/snapshot\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/contract\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/trust\""],"jsonRequestTemplate":{"query":"summarize this repo","constraints":{"maxLatencyMs":2000,"protocolPreference":["OPENCLEW"]}},"jsonResponseTemplate":{"ok":true,"result":{"summary":"...","confidence":0.9},"meta":{"source":"CLAWHUB","generatedAt":"2026-10-09T23:52:04.801Z"}},"retryPolicy":{"maxAttempts":3,"backoffMs":[500,1500,3500],"retryableConditions":["HTTP_429","HTTP_503","NETWORK_TIMEOUT"]}},"endpoints":{"dossierUrl":"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/dossier","snapshotUrl":"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-xucheng-swarm-layer/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":"high","updatedAt":"2026-10-09T20:25:48.962Z","emptyReason":null},"readme":"Skill: Swarm Layer\n\nOwner: xucheng\n\nSummary: OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,...\n\nTags: harness:0.5.12, latest:0.5.12, openclaw:0.5.12, orchestration:0.5.12, swarm:0.5.12, workflow:0.5.12\n\nVersion history:\n\nv0.5.12 | 2026-06-11T14:12:42.520Z | user\n\nDocument artifact acceptance (expectedArtifacts) and the 0.5.12 conservative ACP failure-mapping baseline.\n\nv0.5.11 | 2026-06-05T05:38:54.767Z | user\n\nRepublish the Swarm Layer skill with explicit workdir and current 0.5.11 SKILL.md content.\n\nv0.5.10 | 2026-06-05T05:20:11.389Z | user\n\nRefresh the Swarm Layer skill for ACP runtime-mode diagnostics and current 0.5.10 release metadata.\n\nv0.5.9 | 2026-06-05T03:05:41.271Z | user\n\nRefresh the Swarm Layer skill for npm-project acpx backend discovery and current 0.5.9 release metadata.\n\nv0.5.8 | 2026-06-04T07:29:37.866Z | user\n\nSkill-only correction after the 0.5.7 release: publish the current Daily Papers reliability docs with sync-aware run/status commands and remove stale subagent-era wording.\n\nv0.5.7 | 2026-06-04T07:26:53.830Z | user\n\nRefresh the Swarm Layer skill for Daily Papers-style cron reliability, sync-aware run/status commands, and current release metadata.\n\nv0.5.6 | 2026-05-25T13:56:51.049Z | user\n\nRefresh the Swarm Layer skill for the ClawHub package artifact hotfix and current release metadata.\n\nv0.5.5 | 2026-05-25T13:48:54.635Z | user\n\nRefresh the Swarm Layer skill for the npm 11 install-smoke hotfix and current release metadata.\n\nv0.5.4 | 2026-05-25T13:17:44.597Z | user\n\nRefresh the Swarm Layer skill for autopilot watcher mode, current release metadata, and release validation guidance.\n\nv0.5.3 | 2026-05-04T15:22:07.884Z | user\n\nRefresh the Swarm Layer skill for OpenClaw 2026.5.3-1 compatibility, current release metadata, and startup tool-contract guidance.\n\nv0.5.2 | 2026-04-28T06:04:01.382Z | user\n\nRefresh the Swarm Layer skill for OpenClaw 2026.4.26 compatibility, current release metadata, and mini smoke validation.\n\nv0.5.1 | 2026-04-09T06:38:16.532Z | user\n\nRefresh release metadata after the ACP autopilot service-loop hotfix.\n\nv0.5.0 | 2026-04-09T02:05:55.411Z | user\n\nRefresh skill docs for the autopilot control plane and the manual+ACP runtime surface.\n\nv0.4.0 | 2026-04-07T11:17:24.580Z | user\n\nParallel reliability & retry-recovery (SPEC-003): ACP concurrency protection, reject-retry workflow, parallel dispatch\n\nv0.3.4 | 2026-04-06T08:43:59.548Z | user\n\ntest\n\nv0.3.3 | 2026-03-30T08:05:53.389Z | user\n\nRelease alignment and packaging hygiene patch.\n\nv0.3.1 | 2026-03-27T03:50:16.158Z | user\n\n0.3.1 patch: fix npm-install doctor false negative for ACP public SDK detection.\n\nv0.3.0 | 2026-03-27T03:14:23.003Z | user\n\nACP-first 0.3.0 release: remove ACP bridge, keep legacy subagent opt-in, align docs and manifest.\n\nv0.2.1 | 2026-03-26T06:34:51.640Z | user\n\nAdd OpenClaw 2026.3.24 bridge compatibility, ACP public-export alignment, and refreshed install guidance.\n\nv0.2.0 | 2026-03-25T15:07:06.217Z | user\n\nM4 harness enhancement: GAN-inspired patterns (sprint contracts, evaluator injection, quality rubrics), cross-session continuity, task immutability, session budget, assumption tracking\n\nv0.1.1 | 2026-03-22T12:16:53.781Z | user\n\nAdd feature overview, install paths (ClawHub/npm/GitHub), e2e example, and links\n\nv0.1.0 | 2026-03-22T11:51:31.754Z | user\n\nInitial release: unified skill with setup/operate/diagnose/report/tools modules\n\nArchive index:\n\nArchive v0.5.12: 3 files, 8088 bytes\n\nFiles: skill-card.md (2444b), SKILL.md (16759b), _meta.json (131b)\n\nFile v0.5.12:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.12`, validated on OpenClaw local mini `2026.6.5` with startup tool contracts, plugin doctor checks, conservative ACP failure mapping, artifact-acceptance smoke (positive note-producing task plus negative missing-artifact task on a live ACP backend), unique-bridge-session smoke, parallel admission checks, and the full unit/e2e regression suite.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph, including explicit `[parallel]` phase barriers\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Artifact acceptance** — Optional `expectedArtifacts` (absolute paths or globs) per task; a completed ACP run with missing artifacts is downgraded to `failed` with `failure.source: \"artifact-validation\"` instead of reaching `done`\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware queued-task drain\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /my-project\n\n# 2. Write a spec\ncat > /my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /my-project --spec /my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.12` release was smoke-tested on OpenClaw local mini `2026.6.5` with a Daily Papers-style live ACP workload.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\",\n \"defaultRuntimeMode\": \"auto\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **ACP stuck in read-only** → set `acp.defaultRuntimeMode` to `auto` for Codex-backed workflow tasks\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . --sync # Sync active runs, then see what's ready\nopenclaw swarm run --project . --sync-active --dry-run # Preview after active-run sync\nopenclaw swarm run --project . --sync-active --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --sync-active --parallel 3 # Dispatch up to 3 ready/queued tasks\nopenclaw swarm run --project . --sync-active --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw swarm autopilot pause --project . --reason \"operator review\"\nopenclaw swarm autopilot resume --project .\nopenclaw swarm autopilot tick --project . --dry-run # Inspect next actions without mutating state\nopenclaw swarm autopilot stop --project . --mode graceful\n```\n\nUse autopilot when you want the workflow to keep progressing under policy control instead of manually calling `run`, `session status`, and recovery commands one by one.\n\nOptional watcher mode:\n\n```json\n{\n \"autopilot\": {\n \"enabled\": true,\n \"watcherMode\": \"hybrid\",\n \"watcher\": {\n \"debounceMs\": 100,\n \"safetyTickMs\": 300000,\n \"library\": \"auto\"\n }\n }\n}\n```\n\nUse `watcherMode: \"polling\"` as the rollback path.\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Safe explicit fallback when ACP automation is unavailable or not desired |\n| `acp` | Default-capable automation path through the public ACP control-plane |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Harness Enhancement (GAN-Inspired Patterns)\n\nEnable advanced harness features for long-running agent orchestration:\n\n```json\n{\n \"enforceTaskImmutability\": true,\n \"bootstrap\": { \"enabled\": true },\n \"evaluator\": { \"enabled\": true, \"autoInjectAfter\": [\"coding\"] }\n}\n```\n\n**Sprint Contracts** — Add `Acceptance Criteria` to your spec. `plan` auto-generates a `SprintContract` with verifiable criteria attached to coding tasks.\n\n**Evaluator Injection** — When `evaluator.enabled`, each coding task gets an auto-injected `-eval` task that validates the contract. Dependency chains adjust automatically:\n```\ncoding-task → coding-task-eval → next-task\n```\n\n**Quality Rubrics** — Replace binary approve/reject with 4-dimension weighted scoring:\n- functionality (0.3) / correctness (0.3) / design (0.2) / craft (0.2)\n- Weighted total >= 6.0 → approve; < 6.0 → reject\n\n**Cross-Session Progress** — `progress.json` auto-updated after each `run` and `review`. Bootstrap sequence loads progress on startup: verify env → load progress → select task → verify baseline.\n\n**Task Immutability** — When `enforceTaskImmutability` is enabled, agents cannot mutate task definitions (title, deps, runner, etc.). Only `status`, `review.status`, and `contract.criteria[].passes` are mutable.\n\n**Session Budget** — Set `runner.budget.maxDurationSeconds` and `runner.budget.maxRetries` per task. Exceeded budgets annotate `[BUDGET EXCEEDED]` on run records.\n\n**Assumption Tracking** — `WorkflowState.assumptions` tracks model capability, environment, tooling, and workflow structure assumptions with validation lifecycle.\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"enable harness mode\" | Add evaluator + immutability + bootstrap config |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, ACP readiness failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check ACP readiness and bridge-exit health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Legacy bridge compatibility warning:**\n- `doctor --json` → check `warnings` / `remediation`\n- Common: remove stale `bridge.acpFallbackEnabled`, or clean up old `versionAllow`/`openclawRoot` metadata\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Remove stale bridge metadata if present → rerun doctor → verify runner resolution\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project`, `sync?` | Workflow status with optional active-run sync and attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?`, `parallel?`, `allReady?`, `syncActive?` | Sync active runs when requested, then dispatch runnable or queued tasks |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review (reject returns to ready by default) |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.12:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.12\",\n \"publishedAt\": 1781187162520\n}\n\nFile v0.5.12:skill-card.md\n\n## Description:\n\nSwarm Layer orchestrates spec-driven OpenClaw workflows with task planning, manual or ACP execution, session tracking, review gates, retries, autopilot controls, and reporting.\n\nThis skill is ready for commercial/non-commercial use.\n\n## Publisher:\n\n[xucheng](https://clawhub.ai/user/xucheng)\n\n### License/Terms of Use:\n\nMIT-0\n\n## Use Case:\n\nDevelopers and workflow operators use this skill to plan, run, review, and troubleshoot multi-step OpenClaw project workflows while keeping human review gates and reporting in the loop.\n\n### Deployment Geography for Use:\n\nGlobal\n\n## Known Risks and Mitigations:\n\nRisk: Unpinned external plugin installation paths may resolve to code outside the reviewed 0.5.12 release.\n\nMitigation: Pin installation to the reviewed 0.5.12 release or a specific audited commit, and verify provenance where possible before use.\n\nRisk: Autopilot and watcher modes can continue dispatching workflow actions after they are started.\n\nMitigation: Start with status and dry-run commands, keep autopilot or watcher mode off until the workflow is understood, and review approvals instead of bulk-approving blindly.\n\nRisk: ACP automation can run task workflows in the active project context.\n\nMitigation: Run the skill only in the intended project and fall back to manual execution when ACP readiness is blocked or unclear.\n\n## Reference(s):\n\n- [ClawHub Skill Page](https://clawhub.ai/xucheng/skills/swarm-layer)\n- [OpenClaw Swarm Layer npm Package](https://www.npmjs.com/package/openclaw-swarm-layer)\n- [OpenClaw Swarm Layer Documentation](https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs)\n\n## Skill Output:\n\n**Output Type(s):** [text, markdown, code, shell commands, configuration, guidance]\n\n**Output Format:** [Markdown guidance with inline shell commands, JSON configuration examples, and operational status summaries]\n\n**Output Parameters:** [1D]\n\n**Other Properties Related to Output:** [May create local workflow reports such as status snapshots, run logs, review logs, spec archives, and completion summaries when the referenced plugin workflow is executed.]\n\n## Skill Version(s):\n\n0.5.12 (source: server release metadata)\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\nArchive v0.5.11: 3 files, 8069 bytes\n\nFiles: skill-card.md (2929b), SKILL.md (16523b), _meta.json (131b)\n\nFile v0.5.11:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.11`, validated on OpenClaw package baseline `2026.5.3-1` and local mini `2026.6.1` with startup tool contracts, plugin doctor checks, runtime health smoke, autopilot watcher smoke, Daily Papers-style local smoke, npm-project `acpx` backend discovery, ACP runtime-mode smoke, explicit-workdir ClawHub skill publish checks, and remote package-install smoke.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph, including explicit `[parallel]` phase barriers\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware queued-task drain\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /my-project\n\n# 2. Write a spec\ncat > /my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /my-project --spec /my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.11` release was smoke-tested on OpenClaw package baseline `2026.5.3-1` and local mini `2026.6.1`.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\",\n \"defaultRuntimeMode\": \"auto\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **ACP stuck in read-only** → set `acp.defaultRuntimeMode` to `auto` for Codex-backed workflow tasks\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . --sync # Sync active runs, then see what's ready\nopenclaw swarm run --project . --sync-active --dry-run # Preview after active-run sync\nopenclaw swarm run --project . --sync-active --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --sync-active --parallel 3 # Dispatch up to 3 ready/queued tasks\nopenclaw swarm run --project . --sync-active --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw swarm autopilot pause --project . --reason \"operator review\"\nopenclaw swarm autopilot resume --project .\nopenclaw swarm autopilot tick --project . --dry-run # Inspect next actions without mutating state\nopenclaw swarm autopilot stop --project . --mode graceful\n```\n\nUse autopilot when you want the workflow to keep progressing under policy control instead of manually calling `run`, `session status`, and recovery commands one by one.\n\nOptional watcher mode:\n\n```json\n{\n \"autopilot\": {\n \"enabled\": true,\n \"watcherMode\": \"hybrid\",\n \"watcher\": {\n \"debounceMs\": 100,\n \"safetyTickMs\": 300000,\n \"library\": \"auto\"\n }\n }\n}\n```\n\nUse `watcherMode: \"polling\"` as the rollback path.\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Safe explicit fallback when ACP automation is unavailable or not desired |\n| `acp` | Default-capable automation path through the public ACP control-plane |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Harness Enhancement (GAN-Inspired Patterns)\n\nEnable advanced harness features for long-running agent orchestration:\n\n```json\n{\n \"enforceTaskImmutability\": true,\n \"bootstrap\": { \"enabled\": true },\n \"evaluator\": { \"enabled\": true, \"autoInjectAfter\": [\"coding\"] }\n}\n```\n\n**Sprint Contracts** — Add `Acceptance Criteria` to your spec. `plan` auto-generates a `SprintContract` with verifiable criteria attached to coding tasks.\n\n**Evaluator Injection** — When `evaluator.enabled`, each coding task gets an auto-injected `-eval` task that validates the contract. Dependency chains adjust automatically:\n```\ncoding-task → coding-task-eval → next-task\n```\n\n**Quality Rubrics** — Replace binary approve/reject with 4-dimension weighted scoring:\n- functionality (0.3) / correctness (0.3) / design (0.2) / craft (0.2)\n- Weighted total >= 6.0 → approve; < 6.0 → reject\n\n**Cross-Session Progress** — `progress.json` auto-updated after each `run` and `review`. Bootstrap sequence loads progress on startup: verify env → load progress → select task → verify baseline.\n\n**Task Immutability** — When `enforceTaskImmutability` is enabled, agents cannot mutate task definitions (title, deps, runner, etc.). Only `status`, `review.status`, and `contract.criteria[].passes` are mutable.\n\n**Session Budget** — Set `runner.budget.maxDurationSeconds` and `runner.budget.maxRetries` per task. Exceeded budgets annotate `[BUDGET EXCEEDED]` on run records.\n\n**Assumption Tracking** — `WorkflowState.assumptions` tracks model capability, environment, tooling, and workflow structure assumptions with validation lifecycle.\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"enable harness mode\" | Add evaluator + immutability + bootstrap config |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, ACP readiness failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check ACP readiness and bridge-exit health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Legacy bridge compatibility warning:**\n- `doctor --json` → check `warnings` / `remediation`\n- Common: remove stale `bridge.acpFallbackEnabled`, or clean up old `versionAllow`/`openclawRoot` metadata\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Remove stale bridge metadata if present → rerun doctor → verify runner resolution\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project`, `sync?` | Workflow status with optional active-run sync and attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?`, `parallel?`, `allReady?`, `syncActive?` | Sync active runs when requested, then dispatch runnable or queued tasks |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review (reject returns to ready by default) |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.11:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.11\",\n \"publishedAt\": 1780637934767\n}\n\nFile v0.5.11:skill-card.md\n\n## Description:
\nOpenClaw Swarm Layer turns workflow specifications into executable task graphs with manual fallback, ACP automation, supervised autopilot, persistent sessions, review gates, retries, and operator reporting.
\n\nThis skill is ready for commercial/non-commercial use.
\n\n## Publisher:
\n[xucheng](https://clawhub.ai/user/xucheng)
\n\n### License/Terms of Use:
\nMIT-0
\n\n\n## Use Case:
\nDevelopers and operators use this skill to plan, run, monitor, review, and report multi-step OpenClaw workflows from Markdown specifications. It is suited for local, operator-supervised orchestration where task progress, retries, sessions, and review decisions need to stay visible.
\n\n### Deployment Geography for Use:
\nGlobal
\n\n## Known Risks and Mitigations:
\nRisk: Local reports, logs, progress state, spec archives, and optional Obsidian copies can expose project details or sensitive workflow context.
\nMitigation: Use the skill only in trusted projects, review retention and access for any Obsidian vault before enabling sync, and inspect generated files before sharing them.
\nRisk: Autopilot and watcher modes can continue advancing workflow tasks under policy control after they are started.
\nMitigation: Start with dry-run and manual review, keep autopilot scoped to trusted projects, and use the documented pause, stop, status, and tick controls to supervise execution.
\nRisk: Generated task plans, reviewer guidance, or reports may be incomplete or incorrect for a specific project.
\nMitigation: Keep review gates enabled, approve or reject tasks explicitly, and run project-specific tests or checks before relying on completed work.
\n\n\n## Reference(s):
\n- [ClawHub Skill Page](https://clawhub.ai/xucheng/swarm-layer)
\n- [OpenClaw Swarm Layer GitHub Repository](https://github.com/xucheng/openclaw-swarm-layer)
\n- [OpenClaw Swarm Layer Documentation](https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs)
\n- [npm Package](https://www.npmjs.com/package/openclaw-swarm-layer)
\n\n\n## Skill Output:
\n**Output Type(s):** [text, markdown, shell commands, configuration, guidance]
\n**Output Format:** [Markdown guidance with shell command and JSON configuration snippets, plus local Markdown report files.]
\n**Output Parameters:** [1D]
\n**Other Properties Related to Output:** [The skill can produce local .openclaw swarm status snapshots, run logs, review logs, spec archives, completion summaries, and optional Obsidian copies.]
\n\n## Skill Version(s):
\n0.5.11 (source: server release evidence and SKILL.md release baseline)
\n\n## Ethical Considerations:
\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\nArchive v0.5.10: 3 files, 7857 bytes\n\nFiles: skill-card.md (2518b), SKILL.md (16310b), _meta.json (131b)\n\nFile v0.5.10:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.9`, validated on OpenClaw package baseline `2026.5.3-1` and local mini `2026.6.1` with startup tool contracts, plugin doctor checks, runtime health smoke, autopilot watcher smoke, Daily Papers-style local smoke, npm-project `acpx` backend discovery, and remote package-install smoke.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph, including explicit `[parallel]` phase barriers\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware queued-task drain\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /my-project\n\n# 2. Write a spec\ncat > /my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /my-project --spec /my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.9` release was smoke-tested on OpenClaw package baseline `2026.5.3-1` and local mini `2026.6.1`.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . --sync # Sync active runs, then see what's ready\nopenclaw swarm run --project . --sync-active --dry-run # Preview after active-run sync\nopenclaw swarm run --project . --sync-active --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --sync-active --parallel 3 # Dispatch up to 3 ready/queued tasks\nopenclaw swarm run --project . --sync-active --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw swarm autopilot pause --project . --reason \"operator review\"\nopenclaw swarm autopilot resume --project .\nopenclaw swarm autopilot tick --project . --dry-run # Inspect next actions without mutating state\nopenclaw swarm autopilot stop --project . --mode graceful\n```\n\nUse autopilot when you want the workflow to keep progressing under policy control instead of manually calling `run`, `session status`, and recovery commands one by one.\n\nOptional watcher mode:\n\n```json\n{\n \"autopilot\": {\n \"enabled\": true,\n \"watcherMode\": \"hybrid\",\n \"watcher\": {\n \"debounceMs\": 100,\n \"safetyTickMs\": 300000,\n \"library\": \"auto\"\n }\n }\n}\n```\n\nUse `watcherMode: \"polling\"` as the rollback path.\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Safe explicit fallback when ACP automation is unavailable or not desired |\n| `acp` | Default-capable automation path through the public ACP control-plane |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Harness Enhancement (GAN-Inspired Patterns)\n\nEnable advanced harness features for long-running agent orchestration:\n\n```json\n{\n \"enforceTaskImmutability\": true,\n \"bootstrap\": { \"enabled\": true },\n \"evaluator\": { \"enabled\": true, \"autoInjectAfter\": [\"coding\"] }\n}\n```\n\n**Sprint Contracts** — Add `Acceptance Criteria` to your spec. `plan` auto-generates a `SprintContract` with verifiable criteria attached to coding tasks.\n\n**Evaluator Injection** — When `evaluator.enabled`, each coding task gets an auto-injected `-eval` task that validates the contract. Dependency chains adjust automatically:\n```\ncoding-task → coding-task-eval → next-task\n```\n\n**Quality Rubrics** — Replace binary approve/reject with 4-dimension weighted scoring:\n- functionality (0.3) / correctness (0.3) / design (0.2) / craft (0.2)\n- Weighted total >= 6.0 → approve; < 6.0 → reject\n\n**Cross-Session Progress** — `progress.json` auto-updated after each `run` and `review`. Bootstrap sequence loads progress on startup: verify env → load progress → select task → verify baseline.\n\n**Task Immutability** — When `enforceTaskImmutability` is enabled, agents cannot mutate task definitions (title, deps, runner, etc.). Only `status`, `review.status`, and `contract.criteria[].passes` are mutable.\n\n**Session Budget** — Set `runner.budget.maxDurationSeconds` and `runner.budget.maxRetries` per task. Exceeded budgets annotate `[BUDGET EXCEEDED]` on run records.\n\n**Assumption Tracking** — `WorkflowState.assumptions` tracks model capability, environment, tooling, and workflow structure assumptions with validation lifecycle.\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"enable harness mode\" | Add evaluator + immutability + bootstrap config |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, ACP readiness failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check ACP readiness and bridge-exit health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Legacy bridge compatibility warning:**\n- `doctor --json` → check `warnings` / `remediation`\n- Common: remove stale `bridge.acpFallbackEnabled`, or clean up old `versionAllow`/`openclawRoot` metadata\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Remove stale bridge metadata if present → rerun doctor → verify runner resolution\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project`, `sync?` | Workflow status with optional active-run sync and attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?`, `parallel?`, `allReady?`, `syncActive?` | Sync active runs when requested, then dispatch runnable or queued tasks |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review (reject returns to ready by default) |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.10:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.10\",\n \"publishedAt\": 1780636811389\n}\n\nFile v0.5.10:skill-card.md\n\n## Description:
\nOpenClaw Swarm Layer helps agents turn workflow specifications into executable task graphs, run them through manual or ACP automation, manage persistent sessions and review gates, retry failed work, and report progress.
\n\nThis skill is ready for commercial/non-commercial use.
\n\n## Publisher:
\n[xucheng](https://clawhub.ai/user/xucheng)
\n\n### License/Terms of Use:
\nMIT-0
\n\n\n## Use Case:
\nDevelopers and engineers use this skill to install, configure, operate, diagnose, and report on OpenClaw Swarm Layer workflows for single-project task orchestration with human review gates.
\n\n### Deployment Geography for Use:
\nGlobal
\n\n## Known Risks and Mitigations:
\nRisk: Autopilot and watcher modes can keep a workflow progressing after setup.
\nMitigation: Keep autopilot and watchers disabled until dry-run output, pause/resume/stop controls, and retry limits have been reviewed.
\nRisk: Persistent sessions and workflow reports can record project specs, logs, and review decisions.
\nMitigation: Use the skill only on projects where local workflow state is acceptable, and avoid Obsidian sync for sensitive specs or logs.
\nRisk: Automated ACP execution can dispatch project tasks and write reports.
\nMitigation: Use manual runner mode or dry runs first, and require explicit review approval before marking tasks complete.
\n\n\n## Reference(s):
\n- [ClawHub Skill Page](https://clawhub.ai/xucheng/swarm-layer)
\n- [Source Repository](https://github.com/xucheng/openclaw-swarm-layer)
\n- [Documentation](https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs)
\n- [npm Package](https://www.npmjs.com/package/openclaw-swarm-layer)
\n\n\n## Skill Output:
\n**Output Type(s):** [text, markdown, shell commands, configuration, guidance]
\n**Output Format:** [Markdown with inline shell commands and JSON configuration snippets]
\n**Output Parameters:** [1D]
\n**Other Properties Related to Output:** [Produces operational guidance, command sequences, diagnostic steps, review instructions, and report-oriented summaries for OpenClaw Swarm Layer workflows.]
\n\n## Skill Version(s):
\n0.5.10 (source: server release metadata)
\n\n## Ethical Considerations:
\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\nArchive v0.5.9: 3 files, 4721 bytes\n\nFiles: _meta.json (130b), skill-card.md (2061b), SKILL.md (8282b)\n\nFile v0.5.9:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP/subagent execution, persistent sessions, review gates, and automatic retry. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nUnified skill for spec-driven workflow orchestration. Routes to the appropriate module based on user intent.\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, bridge configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.2.24\n```\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure Bridge (for ACP/subagent execution)\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n },\n \"bridge\": {\n \"enabled\": true,\n \"nodePath\": \"$(which node)\",\n \"openclawRoot\": \"$(npm root -g)/openclaw\",\n \"versionAllow\": [\"CURRENT_VERSION\"]\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"obsidianJournal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **Bridge blocked** → `openclaw swarm doctor --json`, follow `remediation`\n- **Version mismatch** → update `bridge.versionAllow`\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . # See what's ready\nopenclaw swarm run --project . --dry-run # Preview\nopenclaw swarm run --project . --runner acp # Execute (acp/manual/subagent)\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\n```\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Operator executes and records manually (default) |\n| `acp` | Delegate to external harness (Codex, Claude Code) |\n| `subagent` | OpenClaw-native child agent delegation |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, bridge failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check bridge health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Bridge failure:**\n- `doctor --json` → check `blockers` → follow `remediation`\n- Common: update `bridge.versionAllow`, check `nodePath`/`openclawRoot`\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Update bridge.versionAllow → rerun tests → verify doctor\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project` | Workflow status with attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?` | Dispatch next runnable task |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.9:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.9\",\n \"publishedAt\": 1780628741271\n}\n\nFile v0.5.9:skill-card.md\n\n## Description:
\nOpenClaw Swarm Layer supports spec-driven workflow orchestration with ACP/subagent execution, persistent sessions, review gates, automatic retry, setup, operation, diagnosis, and reporting.
\n\nThis skill is ready for commercial/non-commercial use.
\n\n## Publisher:
\n[xucheng](https://clawhub.ai/user/xucheng)
\n\n### License/Terms of Use:
\nMIT-0
\n\n\n## Use Case:
\nDevelopers and engineering teams use this skill to plan, run, review, diagnose, and report OpenClaw swarm workflows. It helps coordinate spec-driven task execution with optional ACP or subagent delegation and persistent sessions.
\n\n### Deployment Geography for Use:
\nGlobal
\n\n## Known Risks and Mitigations:
\nRisk: ACP, subagent execution, or persistent sessions may delegate actions inside a project.
\nMitigation: Enable delegated execution only for projects where the operator trusts the OpenClaw swarm plugin and understands what agents may do.
\nRisk: Workflow reports may be written locally and optionally mirrored into an Obsidian vault.
\nMitigation: Review report contents before enabling sync to shared or cloud-synced vaults, especially for sensitive project workflows.
\n\n\n## Reference(s):
\n- [ClawHub skill page](https://clawhub.ai/xucheng/swarm-layer)
\n\n\n## Skill Output:
\n**Output Type(s):** [text, markdown, shell commands, configuration, guidance]
\n**Output Format:** [Markdown guidance with inline shell and JSON configuration examples]
\n**Output Parameters:** [1D]
\n**Other Properties Related to Output:** [May direct agents to create or read local OpenClaw swarm reports and optional Obsidian mirrors when configured.]
\n\n## Skill Version(s):
\n0.5.9 (source: server-resolved release metadata)
\n\n## Ethical Considerations:
\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\nArchive v0.5.8: 3 files, 7916 bytes\n\nFiles: skill-card.md (2719b), SKILL.md (16272b), _meta.json (130b)\n\nFile v0.5.8:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.7`, validated on OpenClaw package baseline `2026.5.3-1` and local mini `2026.6.1` with startup tool contracts, plugin doctor checks, runtime health smoke, autopilot watcher smoke, Daily Papers-style local smoke, and remote package-install smoke.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph, including explicit `[parallel]` phase barriers\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware queued-task drain\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /my-project\n\n# 2. Write a spec\ncat > /my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /my-project --spec /my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.7` release was smoke-tested on OpenClaw package baseline `2026.5.3-1` and local mini `2026.6.1`.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . --sync # Sync active runs, then see what's ready\nopenclaw swarm run --project . --sync-active --dry-run # Preview after active-run sync\nopenclaw swarm run --project . --sync-active --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --sync-active --parallel 3 # Dispatch up to 3 ready/queued tasks\nopenclaw swarm run --project . --sync-active --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw swarm autopilot pause --project . --reason \"operator review\"\nopenclaw swarm autopilot resume --project .\nopenclaw swarm autopilot tick --project . --dry-run # Inspect next actions without mutating state\nopenclaw swarm autopilot stop --project . --mode graceful\n```\n\nUse autopilot when you want the workflow to keep progressing under policy control instead of manually calling `run`, `session status`, and recovery commands one by one.\n\nOptional watcher mode:\n\n```json\n{\n \"autopilot\": {\n \"enabled\": true,\n \"watcherMode\": \"hybrid\",\n \"watcher\": {\n \"debounceMs\": 100,\n \"safetyTickMs\": 300000,\n \"library\": \"auto\"\n }\n }\n}\n```\n\nUse `watcherMode: \"polling\"` as the rollback path.\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Safe explicit fallback when ACP automation is unavailable or not desired |\n| `acp` | Default-capable automation path through the public ACP control-plane |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Harness Enhancement (GAN-Inspired Patterns)\n\nEnable advanced harness features for long-running agent orchestration:\n\n```json\n{\n \"enforceTaskImmutability\": true,\n \"bootstrap\": { \"enabled\": true },\n \"evaluator\": { \"enabled\": true, \"autoInjectAfter\": [\"coding\"] }\n}\n```\n\n**Sprint Contracts** — Add `Acceptance Criteria` to your spec. `plan` auto-generates a `SprintContract` with verifiable criteria attached to coding tasks.\n\n**Evaluator Injection** — When `evaluator.enabled`, each coding task gets an auto-injected `-eval` task that validates the contract. Dependency chains adjust automatically:\n```\ncoding-task → coding-task-eval → next-task\n```\n\n**Quality Rubrics** — Replace binary approve/reject with 4-dimension weighted scoring:\n- functionality (0.3) / correctness (0.3) / design (0.2) / craft (0.2)\n- Weighted total >= 6.0 → approve; < 6.0 → reject\n\n**Cross-Session Progress** — `progress.json` auto-updated after each `run` and `review`. Bootstrap sequence loads progress on startup: verify env → load progress → select task → verify baseline.\n\n**Task Immutability** — When `enforceTaskImmutability` is enabled, agents cannot mutate task definitions (title, deps, runner, etc.). Only `status`, `review.status`, and `contract.criteria[].passes` are mutable.\n\n**Session Budget** — Set `runner.budget.maxDurationSeconds` and `runner.budget.maxRetries` per task. Exceeded budgets annotate `[BUDGET EXCEEDED]` on run records.\n\n**Assumption Tracking** — `WorkflowState.assumptions` tracks model capability, environment, tooling, and workflow structure assumptions with validation lifecycle.\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"enable harness mode\" | Add evaluator + immutability + bootstrap config |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, ACP readiness failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check ACP readiness and bridge-exit health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Legacy bridge compatibility warning:**\n- `doctor --json` → check `warnings` / `remediation`\n- Common: remove stale `bridge.acpFallbackEnabled`, or clean up old `versionAllow`/`openclawRoot` metadata\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Remove stale bridge metadata if present → rerun doctor → verify runner resolution\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project`, `sync?` | Workflow status with optional active-run sync and attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?`, `parallel?`, `allReady?`, `syncActive?` | Sync active runs when requested, then dispatch runnable or queued tasks |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review (reject returns to ready by default) |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.8:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.8\",\n \"publishedAt\": 1780558177866\n}\n\nFile v0.5.8:skill-card.md\n\n## Description:
\nOpenClaw Swarm Layer turns workflow specifications into executable task graphs with ACP-first execution, supervised autopilot controls, manual fallback, persistent sessions, review gates, retry handling, parallel dispatch, and operator reporting.
\n\nThis skill is ready for commercial/non-commercial use.
\n\n## Publisher:
\n[xucheng](https://clawhub.ai/user/xucheng)
\n\n### License/Terms of Use:
\nMIT-0
\n\n\n## Use Case:
\nDevelopers and engineering operators use this skill to plan, run, review, diagnose, and report on OpenClaw workflow orchestration from Markdown specifications. It helps coordinate ACP or manual execution while keeping human review decisions in the loop.
\n\n### Deployment Geography for Use:
\nGlobal
\n\n## Known Risks and Mitigations:
\nRisk: Automation can execute agent workflows in projects where ACP or autopilot behavior is enabled.
\nMitigation: Review the selected plugin source or package before installation and keep ACP/autopilot scoped to projects where agent execution is acceptable.
\nRisk: Reports, specs, task logs, session history, or review logs may persist locally or be mirrored to Obsidian when configured.
\nMitigation: Avoid enabling Obsidian sync for sensitive workflow data unless the vault location is private and intended for that data.
\nRisk: Workflow proposals or generated task plans may be incomplete or incorrect.
\nMitigation: Use the skill's review gates and operator approval flow before treating task results as complete.
\n\n\n## Reference(s):
\n- [ClawHub skill page](https://clawhub.ai/xucheng/swarm-layer)
\n- [GitHub repository](https://github.com/xucheng/openclaw-swarm-layer)
\n- [Documentation](https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs)
\n- [npm package](https://www.npmjs.com/package/openclaw-swarm-layer)
\n- [Issue tracker](https://github.com/xucheng/openclaw-swarm-layer/issues)
\n\n\n## Skill Output:
\n**Output Type(s):** [guidance, shell commands, configuration, markdown, text]
\n**Output Format:** [Markdown with inline shell commands and JSON configuration examples]
\n**Output Parameters:** [1D]
\n**Other Properties Related to Output:** [May guide creation of local workflow state, run logs, review logs, reports, spec archives, and optional Obsidian-synced copies.]
\n\n## Skill Version(s):
\n0.5.8 (source: server release evidence)
\n\n## Ethical Considerations:
\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\nArchive v0.5.7: 3 files, 4777 bytes\n\nFiles: _meta.json (130b), skill-card.md (2183b), SKILL.md (8282b)\n\nFile v0.5.7:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP/subagent execution, persistent sessions, review gates, and automatic retry. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nUnified skill for spec-driven workflow orchestration. Routes to the appropriate module based on user intent.\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, bridge configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.2.24\n```\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure Bridge (for ACP/subagent execution)\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n },\n \"bridge\": {\n \"enabled\": true,\n \"nodePath\": \"$(which node)\",\n \"openclawRoot\": \"$(npm root -g)/openclaw\",\n \"versionAllow\": [\"CURRENT_VERSION\"]\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"obsidianJournal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **Bridge blocked** → `openclaw swarm doctor --json`, follow `remediation`\n- **Version mismatch** → update `bridge.versionAllow`\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . # See what's ready\nopenclaw swarm run --project . --dry-run # Preview\nopenclaw swarm run --project . --runner acp # Execute (acp/manual/subagent)\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\n```\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Operator executes and records manually (default) |\n| `acp` | Delegate to external harness (Codex, Claude Code) |\n| `subagent` | OpenClaw-native child agent delegation |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, bridge failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check bridge health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Bridge failure:**\n- `doctor --json` → check `blockers` → follow `remediation`\n- Common: update `bridge.versionAllow`, check `nodePath`/`openclawRoot`\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Update bridge.versionAllow → rerun tests → verify doctor\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project` | Workflow status with attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?` | Dispatch next runnable task |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.7:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.7\",\n \"publishedAt\": 1780558013830\n}\n\nFile v0.5.7:skill-card.md\n\n## Description:
\nOpenClaw Swarm Layer: spec-driven workflow orchestration with ACP/subagent execution, persistent sessions, review gates, and automatic retry. Covers setup, operation, diagnosis, and reporting.
\n\nThis skill is ready for commercial/non-commercial use.
\n\n## Publisher:
\n[xucheng](https://clawhub.ai/user/xucheng)
\n\n### License/Terms of Use:
\nMIT-0
\n\n\n## Use Case:
\nDevelopers and operators use this skill to set up and run spec-driven OpenClaw swarm workflows, manage ACP/subagent/manual execution, diagnose stuck or blocked sessions, and summarize workflow status and reports.
\n\n### Deployment Geography for Use:
\nGlobal
\n\n## Known Risks and Mitigations:
\nRisk: Workflow reports may persist project details in local OpenClaw report folders or an optional Obsidian mirror.
\nMitigation: Review report locations before use, avoid including secrets or sensitive details in reviewed material, and delete or redirect generated files when retention is not desired.
\nRisk: ACP/subagent execution and persistent sessions can delegate work to external harnesses and leave sessions active if a run stalls.
\nMitigation: Use doctor, status, dry-run, and review gates before execution; restrict allowed agents and cancel or clean up blocked or stale sessions.
\n\n\n## Reference(s):
\n- [Swarm Layer on ClawHub](https://clawhub.ai/xucheng/swarm-layer)
\n\n\n## Skill Output:
\n**Output Type(s):** [Text, Markdown, Shell commands, Configuration, Guidance]
\n**Output Format:** [Markdown guidance with shell command and JSON configuration examples.]
\n**Output Parameters:** [1D]
\n**Other Properties Related to Output:** [Includes workflow status summaries, review guidance, and report-file paths; local report persistence may include project details.]
\n\n## Skill Version(s):
\n0.5.7 (source: server release metadata)
\n\n## Ethical Considerations:
\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\nArchive v0.5.6: 3 files, 7683 bytes\n\nFiles: skill-card.md (2501b), SKILL.md (15910b), _meta.json (130b)\n\nFile v0.5.6:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.6`, validated on `OpenClaw 2026.5.3-1` with startup tool contracts, plugin doctor checks, runtime health smoke, autopilot watcher smoke, and remote package-install smoke.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware slot management\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /tmp/my-project\n\n# 2. Write a spec\ncat > /tmp/my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /tmp/my-project --spec /tmp/my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /tmp/my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /tmp/my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /tmp/my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /tmp/my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.6` release was smoke-tested on `OpenClaw 2026.5.3-1`.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . # See what's ready\nopenclaw swarm run --project . --dry-run # Preview\nopenclaw swarm run --project . --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --parallel 3 # Dispatch up to 3 ready tasks\nopenclaw swarm run --project . --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw swarm autopilot pause --project . --reason \"operator review\"\nopenclaw swarm autopilot resume --project .\nopenclaw swarm autopilot tick --project . --dry-run # Inspect next actions without mutating state\nopenclaw swarm autopilot stop --project . --mode graceful\n```\n\nUse autopilot when you want the workflow to keep progressing under policy control instead of manually calling `run`, `session status`, and recovery commands one by one.\n\nOptional watcher mode:\n\n```json\n{\n \"autopilot\": {\n \"enabled\": true,\n \"watcherMode\": \"hybrid\",\n \"watcher\": {\n \"debounceMs\": 100,\n \"safetyTickMs\": 300000,\n \"library\": \"auto\"\n }\n }\n}\n```\n\nUse `watcherMode: \"polling\"` as the rollback path.\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Safe explicit fallback when ACP automation is unavailable or not desired |\n| `acp` | Default-capable automation path through the public ACP control-plane |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Harness Enhancement (GAN-Inspired Patterns)\n\nEnable advanced harness features for long-running agent orchestration:\n\n```json\n{\n \"enforceTaskImmutability\": true,\n \"bootstrap\": { \"enabled\": true },\n \"evaluator\": { \"enabled\": true, \"autoInjectAfter\": [\"coding\"] }\n}\n```\n\n**Sprint Contracts** — Add `Acceptance Criteria` to your spec. `plan` auto-generates a `SprintContract` with verifiable criteria attached to coding tasks.\n\n**Evaluator Injection** — When `evaluator.enabled`, each coding task gets an auto-injected `-eval` task that validates the contract. Dependency chains adjust automatically:\n```\ncoding-task → coding-task-eval → next-task\n```\n\n**Quality Rubrics** — Replace binary approve/reject with 4-dimension weighted scoring:\n- functionality (0.3) / correctness (0.3) / design (0.2) / craft (0.2)\n- Weighted total >= 6.0 → approve; < 6.0 → reject\n\n**Cross-Session Progress** — `progress.json` auto-updated after each `run` and `review`. Bootstrap sequence loads progress on startup: verify env → load progress → select task → verify baseline.\n\n**Task Immutability** — When `enforceTaskImmutability` is enabled, agents cannot mutate task definitions (title, deps, runner, etc.). Only `status`, `review.status`, and `contract.criteria[].passes` are mutable.\n\n**Session Budget** — Set `runner.budget.maxDurationSeconds` and `runner.budget.maxRetries` per task. Exceeded budgets annotate `[BUDGET EXCEEDED]` on run records.\n\n**Assumption Tracking** — `WorkflowState.assumptions` tracks model capability, environment, tooling, and workflow structure assumptions with validation lifecycle.\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"enable harness mode\" | Add evaluator + immutability + bootstrap config |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, ACP readiness failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check ACP readiness and bridge-exit health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Legacy bridge compatibility warning:**\n- `doctor --json` → check `warnings` / `remediation`\n- Common: remove stale `bridge.acpFallbackEnabled`, or clean up old `versionAllow`/`openclawRoot` metadata\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Remove stale bridge metadata if present → rerun doctor → verify runner resolution\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project` | Workflow status with attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?`, `parallel?`, `allReady?` | Dispatch runnable tasks (single or batch) |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review (reject returns to ready by default) |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.6:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.6\",\n \"publishedAt\": 1779717411049\n}\n\nFile v0.5.6:skill-card.md\n\n## Description:
\nOpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting.
\n\nThis skill is ready for commercial/non-commercial use.
\n\n## Publisher:
\n[xucheng](https://clawhub.ai/user/xucheng)
\n\n### License/Terms of Use:
\nMIT-0
\n\n\n## Use Case:
\nDevelopers and engineers use this skill to turn workflow specifications into executable OpenClaw task graphs, dispatch tasks, supervise sessions, apply review gates, and generate operational reports.
\n\n### Deployment Geography for Use:
\nGlobal
\n\n## Known Risks and Mitigations:
\nRisk: The skill can plan and dispatch OpenClaw swarm tasks, maintain session state, and manage workflow execution.
\nMitigation: Install and use it only on projects where that authority is acceptable, prefer explicit swarm-related requests, and preview runs with dry-run when possible.
\nRisk: Workflow reports and optional Obsidian mirroring can write project data to local or vault paths.
\nMitigation: Keep reports in appropriate project locations and enable Obsidian mirroring only for vaults that are suitable for the project data.
\n\n\n## Reference(s):
\n- [ClawHub Skill Page](https://clawhub.ai/xucheng/swarm-layer)
\n- [OpenClaw Swarm Layer Package](https://www.npmjs.com/package/openclaw-swarm-layer)
\n- [OpenClaw Swarm Layer Documentation](https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs)
\n- [OpenClaw Swarm Layer Issues](https://github.com/xucheng/openclaw-swarm-layer/issues)
\n\n\n## Skill Output:
\n**Output Type(s):** [text, markdown, shell commands, configuration, guidance]
\n**Output Format:** [Markdown with inline shell commands and configuration snippets]
\n**Output Parameters:** [1D]
\n**Other Properties Related to Output:** [May include local report paths and optional Obsidian mirroring guidance.]
\n\n## Skill Version(s):
\n0.5.6 (source: server release evidence and artifact release baseline)
\n\n## Ethical Considerations:
\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\nArchive v0.5.5: 2 files, 6411 bytes\n\nFiles: SKILL.md (15910b), _meta.json (130b)\n\nFile v0.5.5:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.5`, validated on `OpenClaw 2026.5.3-1` with startup tool contracts, plugin doctor checks, runtime health smoke, autopilot watcher smoke, and remote package-install smoke.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware slot management\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /tmp/my-project\n\n# 2. Write a spec\ncat > /tmp/my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /tmp/my-project --spec /tmp/my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /tmp/my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /tmp/my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /tmp/my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /tmp/my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.5` release was smoke-tested on `OpenClaw 2026.5.3-1`.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . # See what's ready\nopenclaw swarm run --project . --dry-run # Preview\nopenclaw swarm run --project . --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --parallel 3 # Dispatch up to 3 ready tasks\nopenclaw swarm run --project . --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw swarm autopilot pause --project . --reason \"operator review\"\nopenclaw swarm autopilot resume --project .\nopenclaw swarm autopilot tick --project . --dry-run # Inspect next actions without mutating state\nopenclaw swarm autopilot stop --project . --mode graceful\n```\n\nUse autopilot when you want the workflow to keep progressing under policy control instead of manually calling `run`, `session status`, and recovery commands one by one.\n\nOptional watcher mode:\n\n```json\n{\n \"autopilot\": {\n \"enabled\": true,\n \"watcherMode\": \"hybrid\",\n \"watcher\": {\n \"debounceMs\": 100,\n \"safetyTickMs\": 300000,\n \"library\": \"auto\"\n }\n }\n}\n```\n\nUse `watcherMode: \"polling\"` as the rollback path.\n\n### Runner Selection\n| Runner | Use When |\n|--------|----------|\n| `manual` | Safe explicit fallback when ACP automation is unavailable or not desired |\n| `acp` | Default-capable automation path through the public ACP control-plane |\n\n### Session Operations\n```bash\nsession list --project . # List all\nsession inspect --project . --session # Details\nsession follow-up --project . --session --task # Inject task\nsession steer --project . --session --message # Redirect\nsession cancel --project . --run # Abort\nsession cleanup --project . --stale-minutes 60 # Clean orphans\n```\n\n### Session Policies\n| Policy | Behavior |\n|--------|----------|\n| `none` | New session each run (default) |\n| `create_persistent` | Creates reusable persistent session |\n| `reuse_if_available` | Reuse idle persistent session if match found |\n| `require_existing` | Fail if no matching session exists |\n\n### Harness Enhancement (GAN-Inspired Patterns)\n\nEnable advanced harness features for long-running agent orchestration:\n\n```json\n{\n \"enforceTaskImmutability\": true,\n \"bootstrap\": { \"enabled\": true },\n \"evaluator\": { \"enabled\": true, \"autoInjectAfter\": [\"coding\"] }\n}\n```\n\n**Sprint Contracts** — Add `Acceptance Criteria` to your spec. `plan` auto-generates a `SprintContract` with verifiable criteria attached to coding tasks.\n\n**Evaluator Injection** — When `evaluator.enabled`, each coding task gets an auto-injected `-eval` task that validates the contract. Dependency chains adjust automatically:\n```\ncoding-task → coding-task-eval → next-task\n```\n\n**Quality Rubrics** — Replace binary approve/reject with 4-dimension weighted scoring:\n- functionality (0.3) / correctness (0.3) / design (0.2) / craft (0.2)\n- Weighted total >= 6.0 → approve; < 6.0 → reject\n\n**Cross-Session Progress** — `progress.json` auto-updated after each `run` and `review`. Bootstrap sequence loads progress on startup: verify env → load progress → select task → verify baseline.\n\n**Task Immutability** — When `enforceTaskImmutability` is enabled, agents cannot mutate task definitions (title, deps, runner, etc.). Only `status`, `review.status`, and `contract.criteria[].passes` are mutable.\n\n**Session Budget** — Set `runner.budget.maxDurationSeconds` and `runner.budget.maxRetries` per task. Exceeded budgets annotate `[BUDGET EXCEEDED]` on run records.\n\n**Assumption Tracking** — `WorkflowState.assumptions` tracks model capability, environment, tooling, and workflow structure assumptions with validation lifecycle.\n\n### Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"start a new workflow\" | Help write spec → `plan` → `status` |\n| \"run the next task\" | `status` → dry-run → `run` |\n| \"what's happening?\" | `status` → `session status` for running tasks |\n| \"approve everything\" | List review queue → approve each |\n| \"enable harness mode\" | Add evaluator + immutability + bootstrap config |\n| \"something is stuck\" | → [Diagnose](#diagnose) module |\n\n---\n\n# Diagnose\n\n## When to Use\nTasks stuck, ACP readiness failures, sessions not updating, dead letters, orphans.\n\n## Diagnostic Flow\n\n```\n1. openclaw swarm doctor --json → Check ACP readiness and bridge-exit health\n2. openclaw swarm status --project . → Find abnormal tasks/sessions\n3. Investigate specific issue (see below)\n```\n\n### Doctor Severity\n| Severity | Meaning | Action |\n|----------|---------|--------|\n| `healthy` | All good | None |\n| `warning` | Works but risky | Address warnings when convenient |\n| `blocked` | Cannot execute | Follow `remediation` immediately |\n\n### Issue Resolution\n\n**Legacy bridge compatibility warning:**\n- `doctor --json` → check `warnings` / `remediation`\n- Common: remove stale `bridge.acpFallbackEnabled`, or clean up old `versionAllow`/`openclawRoot` metadata\n\n**Stuck running task:**\n```bash\nswarm session status --project . --run # Check if session died\nswarm session cancel --project . --run # Force cancel if hung\n```\n\n**Dead letter task** (retries exhausted):\n- `swarm status` → find `dead_letter` tasks\n- Fix root cause → manually reset task to `ready`\n\n**Orphaned sessions** (stale active):\n```bash\nswarm session cleanup --project . --stale-minutes 60\n```\n\n**Version drift** (after OpenClaw upgrade):\n```bash\nswarm doctor --json\n# Remove stale bridge metadata if present → rerun doctor → verify runner resolution\n```\n\n---\n\n# Report\n\n## When to Use\nCheck progress, generate reports, understand session inventory.\n\n## Quick Check\n```bash\nopenclaw swarm status --project . # Structured summary\nopenclaw swarm report --project . # Full Markdown report\n```\n\n## Report Sections\n| Section | Content |\n|---------|---------|\n| **Attention** | Items needing action: review / blocked / running / dead_letter |\n| **Tasks** | All tasks with current status |\n| **Review Queue** | Tasks awaiting approve/reject |\n| **Highlights** | Notable terminal events (completed/failed/cancelled) |\n| **Recommended Actions** | What to do next |\n| **Recent Runs** | Last 5 runs |\n| **Sessions** | Last 5 sessions with state |\n| **Session Reuse Candidates** | Which tasks can reuse which sessions |\n\n## Report Files\n| File | Trigger | Mode |\n|------|---------|------|\n| `swarm-report.md` | Every operation | Overwrite |\n| `run-log.md` | `swarm run` | Append |\n| `review-log.md` | `swarm review` | Append |\n| `specs/.md` | `swarm plan` | Create once |\n| `completion-summary.md` | All tasks done | Overwrite |\n\nLocal: `/.openclaw/swarm/reports/` (always)\nObsidian: `//` (optional async mirror)\n\n## Conversational Patterns\n| User Says | Do This |\n|-----------|---------|\n| \"what's the status?\" | `status` → summarize counts + attention |\n| \"show me the report\" | `report` → read and present key sections |\n| \"what needs attention?\" | `status` → focus on `attention` array |\n| \"how are sessions?\" | `status` → show session counts + recent |\n\n---\n\n# Tools Reference\n\nFor AI tool calling. All tools accept `--json` for structured output.\n\n| Tool | Parameters | Purpose |\n|------|-----------|---------|\n| `swarm_status` | `project` | Workflow status with attention items |\n| `swarm_task_plan` | `project`, `spec` | Import spec, generate task graph |\n| `swarm_run` | `project`, `task?`, `dryRun?`, `parallel?`, `allReady?` | Dispatch runnable tasks (single or batch) |\n| `swarm_review_gate` | `project`, `task`, `approve?`, `reject?`, `note?` | Approve/reject review (reject returns to ready by default) |\n| `swarm_session_status` | `project`, `run` | Poll session status |\n| `swarm_session_cancel` | `project`, `run`, `reason?` | Cancel session |\n| `swarm_session_close` | `project`, `run`, `reason?` | Close session |\n\nFile v0.5.5:_meta.json\n\n{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.5\",\n \"publishedAt\": 1779716934635\n}\n\nArchive v0.5.4: 2 files, 6401 bytes\n\nFiles: SKILL.md (15880b), _meta.json (130b)\n\nFile v0.5.4:SKILL.md\n\n---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.4`, validated on `OpenClaw 2026.5.3-1` with startup tool contracts, plugin doctor checks, runtime health smoke, and autopilot watcher smoke.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware slot management\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec archives, completion summaries → local + Obsidian sync\n\n## What It Does NOT Do\n\n- Not a distributed multi-node orchestrator — single machine, single project\n- Not a CI/CD pipeline — no git push, PR creation, or deployment automation\n- Not an autonomous PR factory — operator stays in the loop for review decisions\n\n## Install\n\nThree installation paths:\n\n**ClawHub** (skill only):\n```bash\nopenclaw skills install swarm-layer\n```\n\n**ClawHub package** (full plugin):\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer\n```\n\n**npm** (full plugin):\n```bash\nnpm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer\n```\n\n**GitHub** (source):\n```bash\ngit clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\n```\n\nAfter install, verify: `openclaw plugins info openclaw-swarm-layer` should show `Status: loaded`.\n\n## End-to-End Example\n\n```bash\n# 1. Initialize\nopenclaw swarm init --project /tmp/my-project\n\n# 2. Write a spec\ncat > /tmp/my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /tmp/my-project --spec /tmp/my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /tmp/my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /tmp/my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /tmp/my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /tmp/my-project\n```\n\n## Links\n\n- **GitHub**: https://github.com/xucheng/openclaw-swarm-layer\n- **npm**: https://www.npmjs.com/package/openclaw-swarm-layer\n- **Docs**: https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs\n- **Issues**: https://github.com/xucheng/openclaw-swarm-layer/issues\n- **ClawHub**: https://clawhub.ai/skills/swarm-layer\n\n---\n\n## Module Router\n\nDetermine which module to use based on what the user needs:\n\n| User Intent | Module | Key Commands |\n|------------|--------|-------------|\n| Install, configure, initialize | [Setup](#setup) | `plugins install`, `doctor`, `init` |\n| Plan, execute, review, session ops | [Operate](#operate) | `plan`, `run`, `review`, `session *` |\n| Supervised automation control | [Operate](#operate) | `autopilot status/start/pause/resume/stop/tick` |\n| Something broken, stuck, or failing | [Diagnose](#diagnose) | `doctor`, `session status/cancel/cleanup` |\n| Check progress, read reports | [Report](#report) | `status`, `report`, `session list/inspect` |\n\nWhen unsure, start with `openclaw swarm status --project .` to assess the situation.\n\n---\n\n# Setup\n\n## When to Use\nFirst-time install, ACP configuration, project initialization, or config troubleshooting.\n\n## Flow\n\n### 1. Prerequisites\n```bash\nnode --version # >= 22\nopenclaw --version # >= 2026.3.22\n```\n\nThe `0.5.4` release was smoke-tested on `OpenClaw 2026.5.3-1`.\n\n### 2. Install Plugin\n```bash\nopenclaw plugins install clawhub:openclaw-swarm-layer # Published package\nopenclaw plugins install -l /path/to/openclaw-swarm-layer\nopenclaw plugins info openclaw-swarm-layer # Should show Status: loaded\n```\n\n### 3. Configure ACP Public Path\n```json\n{\n \"plugins\": { \"entries\": { \"openclaw-swarm-layer\": { \"config\": {\n \"defaultRunner\": \"auto\",\n \"acp\": {\n \"enabled\": true,\n \"defaultAgentId\": \"codex\",\n \"allowedAgents\": [\"codex\"],\n \"defaultMode\": \"run\"\n }\n }}}}\n}\n```\n\n### 4. Verify\n```bash\nopenclaw swarm doctor --json\n# severity should be \"healthy\" or \"warning\", not \"blocked\"\n```\n\n### 5. Initialize Project\n```bash\nopenclaw swarm init --project .\n```\n\n### 6. Optional: Obsidian Sync + Journal\n```json\n{\n \"obsidianRoot\": \"/path/to/vault/reports\",\n \"journal\": {\n \"enableRunLog\": true,\n \"enableReviewLog\": true,\n \"enableSpecArchive\": true,\n \"enableCompletionSummary\": true\n }\n}\n```\n\n### Setup Troubleshooting\n- **Plugin not loading** → `openclaw plugins info openclaw-swarm-layer`\n- **ACP unavailable** → `openclaw swarm doctor --json`, confirm public ACP export readiness and runner resolution\n- **Legacy bridge config warning** → remove `bridge.acpFallbackEnabled` after confirming ACP public path is ready\n\n---\n\n# Operate\n\n## When to Use\nPlan and execute workflows, manage sessions, complete review cycles.\n\n## Core Loop\n\n```\nWrite Spec → Plan → Status → Run → Poll Session → Review → Repeat\n```\n\n### Write a Spec\n```markdown\n# My Workflow\n## Goals\n- What to achieve\n## Phases\n### Phase 1\n- Task A\n- Task B\n```\n\n### Plan → Run → Review\n```bash\nopenclaw swarm plan --project . --spec SPEC.md # Import and generate tasks\nopenclaw swarm status --project . # See what's ready\nopenclaw swarm run --project . --dry-run # Preview\nopenclaw swarm run --project . --runner acp # Execute (acp/manual)\nopenclaw swarm run --project . --parallel 3 # Dispatch up to 3 ready tasks\nopenclaw swarm run --project . --all-ready # Fill all available concurrency slots\nopenclaw swarm session status --project . --run # Poll until complete\nopenclaw swarm review --project . --task --approve\nopenclaw swarm review --project . --task --reject --retry-now # Reject and force retry\n```\n\n### Autopilot Control Plane\n```bash\nopenclaw swarm autopilot status --project . # Health + last decision\nopenclaw swarm autopilot start --project . # Start supervised loop\nopenclaw\n\nArchive v0.5.3: 2 files, 6281 bytes\n\nFiles: SKILL.md (15557b), _meta.json (130b)","readmeExcerpt":"Skill: Swarm Layer Owner: xucheng Summary: OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,... Tags: harness:0.5.12, latest:0.5.12, openclaw:0.5.12, orchestration:0.5.12, swarm:0.5.12, workflow:0.5.12 Version history: v0.5.12 | 2026-06-11T14:12:42.520Z | user Document artifact acceptance (expectedArtifacts) and","codeSnippets":[],"executableExamples":[{"language":"bash","snippet":"openclaw skills install swarm-layer"},{"language":"bash","snippet":"openclaw plugins install clawhub:openclaw-swarm-layer"},{"language":"bash","snippet":"npm install openclaw-swarm-layer\nopenclaw plugins install -l node_modules/openclaw-swarm-layer"},{"language":"bash","snippet":"git clone https://github.com/xucheng/openclaw-swarm-layer.git\ncd openclaw-swarm-layer && npm install && npm run build\nopenclaw plugins install -l /path/to/openclaw-swarm-layer"},{"language":"bash","snippet":"# 1. Initialize\nopenclaw swarm init --project /my-project\n\n# 2. Write a spec\ncat > /my-project/SPEC.md << 'EOF'\n# Feature Build\n## Goals\n- Implement and test the new feature\n## Phases\n### Implement\n- Write the core logic\n### Test\n- Run unit tests\nEOF\n\n# 3. Plan\nopenclaw swarm plan --project /my-project --spec /my-project/SPEC.md\n# → specId: feature-build, taskCount: 2\n\n# 4. Execute first task\nopenclaw swarm run --project /my-project --runner acp\n# → action: dispatched, runId: implement-task-1-run-...\n\n# 5. Poll until complete\nopenclaw swarm session status --project /my-project --run \n# → status: completed\n\n# 6. Approve\nopenclaw swarm review --project /my-project --task --approve\n# → status: done\n\n# 7. Execute next task, repeat steps 4-6\n\n# 8. View report\nopenclaw swarm report --project /my-project"},{"language":"bash","snippet":"node --version # >= 22\nopenclaw --version # >= 2026.3.22"}],"parameters":null,"dependencies":[],"permissions":[],"extractedFiles":[{"path":"SKILL.md","content":"---\nname: swarm-layer\ndescription: \"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions, review gates, automatic retry, parallel dispatch, reject-retry workflows, ACP concurrency protection, harness patterns, cross-session continuity, and operator reporting. Covers setup, operation, diagnosis, and reporting.\"\n---\n\n# OpenClaw Swarm Layer\n\nTurn workflow specifications into executable task graphs. Dispatch tasks through manual fallback or ACP automation. Supervise execution through an optional autopilot control plane. Track execution via persistent sessions with reuse and thread binding. Gate completion with review approval. Auto-retry on failure. Generate reports to local disk and Obsidian.\n\nCurrent release baseline: `openclaw-swarm-layer@0.5.12`, validated on OpenClaw local mini `2026.6.5` with startup tool contracts, plugin doctor checks, conservative ACP failure mapping, artifact-acceptance smoke (positive note-producing task plus negative missing-artifact task on a live ACP backend), unique-bridge-session smoke, parallel admission checks, and the full unit/e2e regression suite.\n\n## What It Does\n\n- **Spec-driven planning** — Write a Markdown spec with goals and phased tasks → generates a dependency-ordered task graph, including explicit `[parallel]` phase barriers\n- **Multi-runner execution** — Manual (operator-driven safe fallback) and ACP (default-capable automation path)\n- **Session management** — Persistent sessions with binding-key reuse, thread-bound follow-up, and steering messages\n- **Review gates** — Tasks require explicit approve/reject; structured quality rubrics for weighted multi-dimension scoring\n- **Sprint contracts** — Negotiated verifiable acceptance criteria per task with automated evaluator injection (GAN-inspired pattern)\n- **Artifact acceptance** — Optional `expectedArtifacts` (absolute paths or globs) per task; a completed ACP run with missing artifacts is downgraded to `failed` with `failure.source: \"artifact-validation\"` instead of reaching `done`\n- **Cross-session continuity** — Progress summary synthesis, bootstrap startup sequence, harness assumption tracking\n- **Protective guardrails** — Task field immutability guard, session budget control (duration + retries)\n- **Automatic retry** — Configurable per-task retry policy with dead letter tracking for exhausted tasks\n- **Concurrency protection** — ACP session concurrency limits with queued task scheduling (FIFO)\n- **Reject-retry workflow** — Review rejections return tasks to ready for re-run; configurable retry limits\n- **Parallel dispatch** — `--parallel N` and `--all-ready` batch dispatch with concurrency-aware queued-task drain\n- **Autopilot control plane** — Supervised `status/start/pause/resume/stop/tick` flows with lease-backed decisions, optional watcher-driven ticks, and degraded-mode holds\n- **Operator reports** — Status snapshots, run logs, review logs, spec "},{"path":"_meta.json","content":"{\n \"ownerId\": \"kn7acew66bsj3ef4fmc64ab2px7zzf7b\",\n \"slug\": \"swarm-layer\",\n \"version\": \"0.5.12\",\n \"publishedAt\": 1781187162520\n}"},{"path":"skill-card.md","content":"## Description:\n\nSwarm Layer orchestrates spec-driven OpenClaw workflows with task planning, manual or ACP execution, session tracking, review gates, retries, autopilot controls, and reporting.\n\nThis skill is ready for commercial/non-commercial use.\n\n## Publisher:\n\n[xucheng](https://clawhub.ai/user/xucheng)\n\n### License/Terms of Use:\n\nMIT-0\n\n## Use Case:\n\nDevelopers and workflow operators use this skill to plan, run, review, and troubleshoot multi-step OpenClaw project workflows while keeping human review gates and reporting in the loop.\n\n### Deployment Geography for Use:\n\nGlobal\n\n## Known Risks and Mitigations:\n\nRisk: Unpinned external plugin installation paths may resolve to code outside the reviewed 0.5.12 release.\n\nMitigation: Pin installation to the reviewed 0.5.12 release or a specific audited commit, and verify provenance where possible before use.\n\nRisk: Autopilot and watcher modes can continue dispatching workflow actions after they are started.\n\nMitigation: Start with status and dry-run commands, keep autopilot or watcher mode off until the workflow is understood, and review approvals instead of bulk-approving blindly.\n\nRisk: ACP automation can run task workflows in the active project context.\n\nMitigation: Run the skill only in the intended project and fall back to manual execution when ACP readiness is blocked or unclear.\n\n## Reference(s):\n\n- [ClawHub Skill Page](https://clawhub.ai/xucheng/skills/swarm-layer)\n- [OpenClaw Swarm Layer npm Package](https://www.npmjs.com/package/openclaw-swarm-layer)\n- [OpenClaw Swarm Layer Documentation](https://github.com/xucheng/openclaw-swarm-layer/tree/main/docs)\n\n## Skill Output:\n\n**Output Type(s):** [text, markdown, code, shell commands, configuration, guidance]\n\n**Output Format:** [Markdown guidance with inline shell commands, JSON configuration examples, and operational status summaries]\n\n**Output Parameters:** [1D]\n\n**Other Properties Related to Output:** [May create local workflow reports such as status snapshots, run logs, review logs, spec archives, and completion summaries when the referenced plugin workflow is executed.]\n\n## Skill Version(s):\n\n0.5.12 (source: server release metadata)\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."}],"languages":[],"docsSourceLabel":"CLAWHUB","editorialOverview":"OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,... Skill: Swarm Layer Owner: xucheng Summary: OpenClaw Swarm Layer: spec-driven workflow orchestration with ACP-first execution, supervised autopilot control plane, manual fallback, persistent sessions,... Tags: harness:0.5.12, latest:0.5.12, openclaw:0.5.12, orchestration:0.5.12, swarm:0.5.12, workflow:0.5.12 Version history: v0.5.12 | 2026-06-11T14:12:42.520Z | user Document artifact acceptance (expectedArtifacts) and","editorialQuality":{"score":100,"threshold":65,"status":"ready","wordCount":1206,"uniquenessScore":52,"reasons":[]}},"media":{"evidence":{"source":"no-media","verified":false,"confidence":"low","updatedAt":"2026-10-09T20:25:48.962Z","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":"2026-10-09T20:25:48.962Z","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":"2026-10-09T23:52:04.804Z","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"}]}}}