{"id":"8d2d99c0-d7f9-4e78-969b-46f72c5c724d","entityType":"agent","slug":"clawhub-mikewongonline-autofix","name":"OpenClaw Problem Solver自动修复小龙虾","canonicalUrl":"https://www.xpersona.co/agent/clawhub-mikewongonline-autofix","canonicalPath":"/agent/clawhub-mikewongonline-autofix","generatedAt":"2026-10-13T05:33:18.646Z","source":"CLAWHUB","claimStatus":"UNCLAIMED","verificationTier":"NONE","summary":{"evidence":{"source":"editorial-content","verified":true,"confidence":"high","updatedAt":"2026-10-11T21:21:41.392Z","emptyReason":null},"description":"A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates... Skill: OpenClaw Problem Solver自动修复小龙虾 Owner: mikewongonline Summary: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates... Tags: latest:6.1.0 Version history: v6.1.0 | 2026-05-22T10:24:18.720Z | user **Gateway Watchdog reliability and notification pipeline improvements.** - Gateway Watchdog upgraded to v6.1 with","descriptionLabel":"Technical summary","evidenceSummary":"Capability contract not published. No trust telemetry is available yet. 991 downloads reported by the source. Last updated 10/11/2026.","installCommand":"clawhub skill install s17es3rrg0jybezhyzn66x6tyd83hv2p:autofix","sourceUrl":"https://clawhub.ai/mikewongonline/autofix","homepage":"https://clawhub.ai/mikewongonline/skills/autofix","primaryLinks":[{"label":"View on ClawHub","url":"https://clawhub.ai/mikewongonline/autofix","kind":"source"},{"label":"Homepage","url":"https://clawhub.ai/mikewongonline/skills/autofix","kind":"homepage"}],"safetyScore":84,"overallRank":62,"popularityScore":60,"trustScore":null,"claimedByName":null,"isOwner":false,"seoDescription":"A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates... "},"coverage":{"evidence":{"source":"public-profile","verified":false,"confidence":"medium","updatedAt":"2026-10-11T21:21:41.392Z","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":"2026-10-11T21:21:41.392Z","emptyReason":null},"stars":null,"forks":null,"downloads":991,"likes":null,"task":null,"library":null,"packageName":null,"latestVersion":"6.1.0","tractionLabel":"991 downloads"},"release":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"medium","updatedAt":"2026-10-11T21:21:41.380Z","emptyReason":null},"lastUpdatedAt":"2026-10-11T21:21:41.392Z","lastCrawledAt":"2026-10-11T21:21:41.380Z","lastIndexedAt":null,"nextCrawlAt":"2026-10-12T21:21:41.380Z","lastVerifiedAt":null,"highlights":[{"version":"6.1.0","createdAt":"2026-05-22T10:24:18.720Z","changelog":"**Gateway Watchdog reliability and notification pipeline improvements.** - Gateway Watchdog upgraded to v6.1 with several fixes and enhancements. - Feishu direct message (DM) is now the primary alert channel (instant delivery); WebChat moved to secondary (async, background thread with 60s timeout). - Fixed config cache bug causing `load_gateway_config()` to return `token=None` on cache hit. - Fixed async WebChat thread to prevent main process blocking during model loading. - Error notification now includes full stack traces in Feishu and WebChat alerts. - Updated deployment and prerequisites for easier setup and improved reliability.","fileCount":28,"zipByteSize":109357},{"version":"6.0.0","createdAt":"2026-05-22T08:38:54.213Z","changelog":"autofix v6.0.0 introduces unified runtime health checks, API key validation, improved reporting, and a visual dashboard. - Added runtime health check module (`scripts/runtime_health_check.py`) for comprehensive, scriptable environment inspection. - Integrated API key validation and resource monitoring (`scripts/api_key_validator.py`) to detect misconfigurations and credential issues. - Introduced unified diagnosis formatter (`scripts/diagnosis_formatter.py`) to combine health, doctor, and key status into a single, severity-sorted report and enable regression (before/after) checks. - Launched an interactive health dashboard (`scripts/health_dashboard.py`) for visualizing diagnostic results. - Enhanced documentation with multi-layered diagnosis and regression workflows.","fileCount":27,"zipByteSize":103635},{"version":"5.5.1","createdAt":"2026-05-20T11:28:28.106Z","changelog":"No changes detected in this release. - Version 5.5.1 does not introduce any updates or modifications to the skill files.","fileCount":21,"zipByteSize":64916},{"version":"5.5.0","createdAt":"2026-05-20T11:18:24.280Z","changelog":"**autofix v5.5.0 Changelog** - Introduced Gateway Watchdog (v4.1): An independent background daemon for real-time monitoring and alerting of Gateway health with severity levels, noise filtering, dual-channel notifications, and auto-restart options. - Added auto-repair scripts and logic for common Gateway issues, with safe/verified repair actions based on diagnostic events. - New/updated scripts: `scripts/auto_repair.py`, `scripts/requirements.txt`, `scripts/watchdog_monitor.py`. - Improved documentation: English and Chinese READMEs, streamlined install guides, and clearer privacy/security policy. - Reorganized and expanded quick start and usage files; removed previous EXAMPLE and QUICK_START documents to reflect the new version structure.","fileCount":21,"zipByteSize":64758},{"version":"5.0.0","createdAt":"2026-05-18T06:01:09.484Z","changelog":"Version 5.0.0 introduces major enhancements: - Adds interactive Diagnosis Report Visualization with visual risk flags, evidence diagrams, and rollback command previews. - Launches Error Log Intelligent Summary (ELIS) for automated, LLM-powered root cause analysis and recommended fixes. - Updates core workflow with a new resources pre-check, enhanced 3-step user confirmation before fixes, and stricter privacy handling. - Reorganizes all documentation and module guides for easier discovery and usage. - Includes new scripts for ELIS analysis and visual report generation.","fileCount":17,"zipByteSize":56741}]},"execution":{"evidence":{"source":"CLAWHUB","verified":false,"confidence":"low","updatedAt":null,"emptyReason":"No published capability contract is available yet."},"installCommand":"clawhub skill install s17es3rrg0jybezhyzn66x6tyd83hv2p:autofix","setupComplexity":"medium","setupSteps":["Python environment detected. Create a strict virtual environment (`python -m venv .venv`) before installing dependencies to prevent system-level package conflicts.","Setup complexity is LOW. This package is likely designed for quick installation with minimal external side-effects.","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-mikewongonline-autofix/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/trust"},"curlExamples":["curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/snapshot\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/contract\"","curl -s \"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/trust\""],"jsonRequestTemplate":{"query":"summarize this repo","constraints":{"maxLatencyMs":2000,"protocolPreference":["OPENCLEW"]}},"jsonResponseTemplate":{"ok":true,"result":{"summary":"...","confidence":0.9},"meta":{"source":"CLAWHUB","generatedAt":"2026-10-13T05:33:18.644Z"}},"retryPolicy":{"maxAttempts":3,"backoffMs":[500,1500,3500],"retryableConditions":["HTTP_429","HTTP_503","NETWORK_TIMEOUT"]}},"endpoints":{"dossierUrl":"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/dossier","snapshotUrl":"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/snapshot","contractUrl":"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/contract","trustUrl":"https://www.xpersona.co/api/v1/agents/clawhub-mikewongonline-autofix/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":"2026-10-11T21:21:41.392Z","emptyReason":null},"readme":"Skill: OpenClaw Problem Solver自动修复小龙虾\n\nOwner: mikewongonline\n\nSummary: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates...\n\nTags: latest:6.1.0\n\nVersion history:\n\nv6.1.0 | 2026-05-22T10:24:18.720Z | user\n\n**Gateway Watchdog reliability and notification pipeline improvements.**\n\n- Gateway Watchdog upgraded to v6.1 with several fixes and enhancements.\n- Feishu direct message (DM) is now the primary alert channel (instant delivery); WebChat moved to secondary (async, background thread with 60s timeout).\n- Fixed config cache bug causing `load_gateway_config()` to return `token=None` on cache hit.\n- Fixed async WebChat thread to prevent main process blocking during model loading.\n- Error notification now includes full stack traces in Feishu and WebChat alerts.\n- Updated deployment and prerequisites for easier setup and improved reliability.\n\nv6.0.0 | 2026-05-22T08:38:54.213Z | user\n\nautofix v6.0.0 introduces unified runtime health checks, API key validation, improved reporting, and a visual dashboard.\n\n- Added runtime health check module (`scripts/runtime_health_check.py`) for comprehensive, scriptable environment inspection.\n- Integrated API key validation and resource monitoring (`scripts/api_key_validator.py`) to detect misconfigurations and credential issues.\n- Introduced unified diagnosis formatter (`scripts/diagnosis_formatter.py`) to combine health, doctor, and key status into a single, severity-sorted report and enable regression (before/after) checks.\n- Launched an interactive health dashboard (`scripts/health_dashboard.py`) for visualizing diagnostic results.\n- Enhanced documentation with multi-layered diagnosis and regression workflows.\n\nv5.5.1 | 2026-05-20T11:28:28.106Z | user\n\nNo changes detected in this release.\n\n- Version 5.5.1 does not introduce any updates or modifications to the skill files.\n\nv5.5.0 | 2026-05-20T11:18:24.280Z | user\n\n**autofix v5.5.0 Changelog**\n\n- Introduced Gateway Watchdog (v4.1): An independent background daemon for real-time monitoring and alerting of Gateway health with severity levels, noise filtering, dual-channel notifications, and auto-restart options.\n- Added auto-repair scripts and logic for common Gateway issues, with safe/verified repair actions based on diagnostic events.\n- New/updated scripts: `scripts/auto_repair.py`, `scripts/requirements.txt`, `scripts/watchdog_monitor.py`.\n- Improved documentation: English and Chinese READMEs, streamlined install guides, and clearer privacy/security policy.\n- Reorganized and expanded quick start and usage files; removed previous EXAMPLE and QUICK_START documents to reflect the new version structure.\n\nv5.0.0 | 2026-05-18T06:01:09.484Z | user\n\nVersion 5.0.0 introduces major enhancements:\n\n- Adds interactive Diagnosis Report Visualization with visual risk flags, evidence diagrams, and rollback command previews.\n- Launches Error Log Intelligent Summary (ELIS) for automated, LLM-powered root cause analysis and recommended fixes.\n- Updates core workflow with a new resources pre-check, enhanced 3-step user confirmation before fixes, and stricter privacy handling.\n- Reorganizes all documentation and module guides for easier discovery and usage.\n- Includes new scripts for ELIS analysis and visual report generation.\n\nArchive index:\n\nArchive v6.1.0: 28 files, 109357 bytes\n\nFiles: docs/COMPLETENESS_SUMMARY.md (4057b), docs/enhancement/MODULE_03_Enhancement_Reports.md (10779b), docs/MODULE_01_PreCheck.md (10913b), docs/MODULE_02_SearchChain.md (3526b), docs/MODULE_03_ValidationAction.md (4061b), docs/MODULE_04_Finalization.md (3313b), docs/reports/AUTOFIX_V5.0_SUMMARY.md (7052b), docs/reports/AUTOFIX_V6.0_PROPOSAL.md (9188b), docs/reports/CHANGES_v5.0.md (9233b), docs/reports/CHANGES_v6.1.md (5910b), docs/reports/VERIFICATION_FINAL.md (9972b), docs/tutorials/EXAMPLE_usage.md (14432b), docs/tutorials/QUICK_START_v5.0.md (9944b), INSTALL.md (5930b), README_CN.md (7195b), README.md (7360b), scripts/api_key_validator.py (25349b), scripts/auto_repair.py (9888b), scripts/canvas_report_generator.py (14916b), scripts/diagnosis_formatter.py (21178b), scripts/elis_helper.py (3396b), scripts/health_dashboard.py (12938b), scripts/requirements.txt (27b), scripts/runtime_health_check.py (20853b), scripts/watchdog_monitor.py (28147b), skill-card.md (2470b), SKILL.md (17610b), _meta.json (126b)\n\nFile v6.1.0:SKILL.md\n\n---\nname: autofix\ndescription: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates Proactive Prediction (L2), Robustness Checks (L1), Knowledge Creation (L3), Diagnosis Report Visualization (v5.5), v6.0 Runtime Health + Key Validation + Unified Report + Health Dashboard, and Gateway Watchdog (v6.1).\n---\n\n# 🧠 OpenClaw Problem Solver (v6.0-M4) — Master Document\n\nThis skill acts as an advanced diagnostic, resolution, and validation engine for any question or bug report related to the OpenClaw framework itself. **v6.0 adds four layers: Runtime Health Check (M1), API Key Validation + Resource Monitor (M2), Unified Diagnosis Report + Regression Check (M3), and Interactive Health Dashboard (M4).**\n\n## 🔒 Privacy-First Constraint\n\nAll knowledge storage (memory, logs) and final reports **must** follow these rules:\n1. **API Keys**: Never store API keys in plain text. Use placeholders or hashes (e.g., `sk-********************`).\n2. **Private Details**: Redact sensitive project names, internal credentials, and user identity info unless explicitly authorized.\n\n## 🎯 When to Use This Skill\n\nUse this skill when the user:\n- Asks \"Why is [feature] not working in OpenClaw?\"\n- Reports a specific bug (e.g., \"The `gateway` tool fails with error X\").\n- Needs guidance on how to implement a specific feature using OpenClaw's architecture or tools.\n- Wants to know the best practice for a certain task within the OpenClaw ecosystem, but requires verification.\n\n## 🌟 Recommended Quick Fix Flow (The Golden Path)\n\nFor the vast majority of OpenClaw issues, this sequence provides the fastest path to resolution. Always suggest this flow first when a user reports an unspecified problem or bug!\n\n1. **Unified Diagnosis (一键诊断, v6.0-M3):** Run `python scripts/diagnosis_formatter.py` which auto-collects all three sources (`openclaw doctor` + `runtime_health_check` + `api_key_validator`) into one severity-sorted report.\n2. **Visual Dashboard (可视化仪表盘, v6.0-M4):** Run `python scripts/health_dashboard.py --canvas` to render the report as an interactive HTML dashboard (embed with `[embed ref=\"health_dashboard\" height=\"740\"]`).\n3. **Save Baseline (保存基线, v6.0-M3):** Run `python scripts/diagnosis_formatter.py --save-baseline` before making any fix.\n4. **Resolution Attempt（修复）:** If the report reveals problems, run `openclaw doctor --fix` or apply suggested fixes manually.\n5. **Regression Check (回归验证, v6.0-M3):** After fixes, run `python scripts/diagnosis_formatter.py --compare` to validate what was fixed, what's new, and what's unchanged.\n\n## 🚀 The Evolved Workflow (6-Step Cycle + Proactive Layers)\n\nThe skill operates by strictly following these steps in sequence, enhanced by proactive layers:\n\n## 🤖 Gateway Watchdog (v6.1) — Proactive Stability Layer\n\n**Overview:**\nA background daemon that periodically polls the Gateway health status. Runs independently from user requests, providing real-time monitoring for anomalies such as Gateway downtime, RPC failures, and configuration drift.\n\n**v6.1 Feature Highlights:**\n\n| Feature | Description |\n|---------|-------------|\n| 🎯 Real Health Check | Calls `openclaw gateway status --json`, parses `service.runtime.status` + `rpc.ok` |\n| 🔇 Noise Filtering | Alerts only after ≥3 consecutive failures; resets after ≥3 consecutive successes |\n| 📊 Severity Levels | Four-tier classification (🟢/🟡/🟠/🔴) with auto-escalation |\n| 📡 Dual-Channel Alerting | **Feishu DM (instant, primary)** + WebChat (async thread, secondary) |\n| 🔄 Single Instance | Windows Mutex ensures only one daemon runs at a time |\n| 📦 Log Rotation | Auto-rotates at 5MB, keeps 3 backup files |\n| ⏰ Precise Scheduling | Fixed-minute schedule eliminates cumulative drift |\n| 🔐 Hot-Reload Config | Monitors `openclaw.json` changes and reloads automatically |\n| 🖥️ Auto-Start | Registers in `HKCU\\Run` for auto-launch on user login |\n| 👋 Startup Confirmation | Sends status to both channels on startup |\n| 🐛 **Config Cache Fix** | **Fixed** `load_gateway_config()` returning `token=None` on cache hit (v6.1) |\n| ⏱️ **Async WebChat** | **Fixed** background thread with 60s timeout for model loading (v6.1) |\n| 📝 **Detailed Error Logs** | **Fixed** full stack traces in Feishu + WebChat notifications (v6.1) |\n\n**📊 Severity & Alert Rules:**\n\n| Consecutive Failures | Level | Behavior |\n|---------------------|-------|----------|\n| < 2 | 🟢 Level 1 — Normal | Silent, no notification |\n| 2 | 🟡 Level 2 — Notice | Silent, continue monitoring |\n| ≥ 3 | 🟠 Level 3 — Warning | **Trigger notification** (first time) |\n| ≥ 5 | 🔴 Level 4 — Critical | **Trigger notification** + repeat every 5 failures |\n| Gateway stopped | 🔴 Level 4 — Critical | **Immediate notification** |\n| Recovered for 3 cycles | ✅ Recovered | Send recovery notification |\n\n**🚨 Notification Triggers:**\n1. Severity **escalation** (e.g., 1→3): sends alert\n2. First time hitting **alert threshold** (≥3 consecutive failures): sends alert\n3. At critical level, **every 5 failures**: sends reminder\n4. **System recovery** (abnormal→normal for 3 consecutive checks): sends recovery notice\n\n### Prerequisites\n\n1. **Feishu channel configured in OpenClaw** — `openclaw channels add feishu`\n2. **Environment variable** — Set `WATCHDOG_FEISHU_USER_ID` to your Feishu open_id:\n ```powershell\n $env:WATCHDOG_FEISHU_USER_ID = \"ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\"\n ```\n3. **Gateway HTTP API** (optional, for WebChat channel) — \n ```bash\n openclaw config set gateway.http.endpoints.chatCompletions.enabled true\n openclaw gateway restart\n ```\n\n> ⚠️ **WebChat timeout**: The model inference takes ~40s on first load.\n> The watchdog uses a **background thread with 60s timeout** so it doesn't block the main monitoring loop.\n\n### Deployment (v6.1)\n\nRun the Watchdog as a standalone background process:\n\n```powershell\n# Start\npython scripts\\watchdog_monitor.py\n\n# Install auto-start (launches on user login)\npython scripts\\watchdog_monitor.py --install\n\n# Remove auto-start\npython scripts\\watchdog_monitor.py --uninstall\n```\n\nOr use `Start-Process` for a hidden window:\n\n```powershell\n$py = (Get-Command python).Source\n$script = \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_monitor.py\"\nStart-Process -FilePath $py -ArgumentList $script `\n -WindowStyle Hidden `\n -WorkingDirectory \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\"\n```\n\nCheck process status:\n\n```powershell\nGet-WmiObject Win32_Process -Filter \"Name like 'python%'\" |\n Where-Object { $_.CommandLine -match 'watchdog_monitor' } |\n Select-Object ProcessId, @{n=\"Start\";e={$_.CreationDate}}\n```\n\nStop the Watchdog:\n\n```powershell\n# Find the PID first, then\nStop-Process -Id -Force\n```\n\nView live logs:\n\n```powershell\nGet-Content \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\gateway_watchdog.log\" -Tail 10 -Wait\n```\n\nView state file:\n\n```powershell\nGet-Content \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_state.json\" -Raw | ConvertFrom-Json\n```\n\n\n### 🔗 Notification Architecture (v6.1)\n\n```\nWatchdog (background daemon, 60s interval)\n │\n ├─ [Channel A — PRIMARY] openclaw message send --channel feishu\n │ → Feishu direct message (ou_xxx)\n │ → **Instant delivery, zero token cost**\n │ → Includes full error stack traces\n │\n ├─ [Channel B — SECONDARY] Gateway HTTP API (/v1/chat/completions)\n │ → WebChat live session (agent:main:main)\n │ → **Async background thread** (doesn't block monitoring)\n │ → 60s timeout for model loading (~40s typical)\n │ → token cost: minimal (max_tokens=10)\n │\n └─ [Log] watchdog_state.json (local check history, last 1440)\n gateway_watchdog.log (rotating, 5MB)\n```\n\n**Channel priority:** Feishu is now the primary channel (instant, reliable via CLI).\nWebChat is secondary (async thread, requires model inference).\n\n### Dual-Channel Alerting + Autofix Triggers (v6.1)\n\n**Channel priority has changed in v6.1:**\n- **Feishu (instant)** is now the primary notification channel\n- **WebChat (async thread)** is the secondary channel\n\n- When a WebChat alert arrives (~40s after error), reply with any of these commands to start diagnosis:\n - `run autofix self-check`\n - `check what's wrong with Gateway`\n - `auto repair`\n- Feishu messages serve as the **instant primary** notification (not offline backup)\n- Each alert message includes detailed error context and stack traces\n\n### Workflow Integration + Autofix Linkage\n\nThe Watchdog forms a **Proactive Stability Layer**, independent of the standard diagnostic flow (Steps 0-5). When an anomaly is detected:\n\na. The daemon logs the event and generates a System Health Warning (SHW) report\nb. Sends a real-time alert (with diagnostic guidance + context JSON)\nc. **Auto-repair** low-risk known issues (e.g., CLI path problems) automatically, then verifies\nd. High-risk operations only provide repair suggestions, awaiting user confirmation\n\n## 🛠️ Auto-Repair Module (v1.0)\n\n**Repair Script Library:** `scripts/auto_repair.py`\n\nMatches repair plans based on the diagnostic context from Watchdog alerts:\n\n| Issue | Match Condition | Repair Action | Risk |\n|-------|----------------|---------------|------|\n| Gateway stopped | `status: stopped` | Restart Gateway | 🟡 Needs confirmation |\n| RPC connection failed | `rpc_ok: false` | Restart Gateway | 🟡 Needs confirmation |\n| CLI unavailable | `status: cli_error` | Check installation path | 🟢 Auto-execute |\n| HTTP unreachable | `status: unreachable` | Check port + restart | 🟡 Needs confirmation |\n\n**Repair Verification Loop:**\n1. After auto-repair, wait 3 seconds then re-run health check\n2. Verification passed → send \"Auto-repair succeeded\" confirmation\n3. Verification failed → send \"Still needs manual diagnosis\" escalation\n\n**Health Trend Tracking:**\n- `watchdog_state.json` retains the last 1440 check records (24 hours)\n- Each record includes: timestamp, health status, severity level, source\n- Trend data can be visualized via Canvas health dashboard\n\n### Current Status (v6.1)\n\n- ✅ Real health check — `service.runtime.status = running` + `rpc.ok = true`\n- ✅ Noise filtering — ≥3 failures to trigger, ≥3 successes to reset\n- ✅ Severity levels — Four-tier (🟢/🟡/🟠/🔴) with auto-escalation\n- ✅ Feishu channel (PRIMARY) — `openclaw message send --channel feishu`, zero token cost, instant delivery\n- ✅ WebChat channel (SECONDARY) — Gateway HTTP API `/v1/chat/completions`, async background thread, 60s timeout\n- ✅ Single instance — Windows Mutex prevents duplicates\n- ✅ Log rotation — 5MB auto-rotate, 3 backups\n- ✅ Precise scheduling — Fixed-minute schedule, no drift\n- ✅ Hot-reload config — Watches `openclaw.json` for changes\n- ✅ Auto-start — `HKCU\\Run` registry, launches on user login\n- ✅ Startup confirmation — Sends status to both channels on start\n- ✅ **`--status` command** — Shows real-time state and exits cleanly (no longer starts daemon by accident)\n- ✅ **Stale process cleanup** — Auto-kills orphaned `--status` processes on daemon startup\n- ✅ **Config cache fix** — `load_gateway_config()` no longer returns `token=None` on subsequent calls\n- ✅ **Detailed error logging** — Full stack traces in all notification channels\n- ✅ **Async WebChat delivery** — Background thread prevents blocking main monitor loop\n- 📁 State file: `scripts/watchdog_state.json`\n- 📁 Log file: `scripts/gateway_watchdog.log`\n\n---\n\n## Standard Workflow (6-Step Cycle + Proactive Layers)\n\nThis skill strictly follows these steps in sequence, enhanced by proactive layers:\n\n### Step 0: Resource Pre-check & Cost Management *(New)* — Starting Point\n\nBefore any resource-intensive external search or service call, proactively check API quotas, rate limits, and budget consumption for the current active session. If quota-low alerts or known rate-limit thresholds are hit, pause all execution steps and notify the user with a clear \"resource warning,\" requesting they wait or switch to a low-cost / local alternative.\n\n### Step 1: Primary Search *(See `docs/MODULE_02_SearchChain.md` — Step 1)*\n- Search the official documentation (`docs.openclaw.ai`) for official solutions\n- Gather context information related to the problem\n- Extract key error messages and configuration status\n\n### Step 2: Backup Search *(See `docs/MODULE_02_SearchChain.md` — Step 2)*\n- If the official docs don't provide an answer, search GitHub Issues\n- Look for community-reported problems and solutions\n- Collect code verification requirements or pattern-matching information\n\n### Step 3: Analysis & Decision *(See `docs/MODULE_03_ValidationAction.md` — Step 3)*\n- Choose the best action path based on search results\n- Perform **evidence chain analysis (L1)** to evaluate solution reliability\n- Decide between direct answer, code verification, or contextual inquiry\n\n### Step 4: Validation & Action (v5.0 Enhanced) *(See `docs/MODULE_03_ValidationAction.md` — Step 4 + `docs/MODULE_03_Enhancement_Reports.md`)*\n- Execute validation (MRE) or propose a contextual inquiry\n- Generate an **interactive diagnosis report** (if MRE fails)\n- ✅ **Three-step confirmation before fixes**: Before running any command with system-modifying or wide-ranging effects (e.g., `openclaw doctor --fix`, `exec`/`write`), follow these safety steps:\n 1. **Problem location & explanation**: Explain the diagnosis result and the core issue to be fixed\n 2. **Scope confirmation**: Ask about and record the specific target or runtime environment (e.g., \"This change will only affect the local development config. Do you agree?\")\n 3. **Rollback plan**: Provide an executable one-click rollback command. Only proceed after the user agrees via `/approve`\n\n### Step 5: Finalization & Memory Update *(See `docs/MODULE_04_Finalization.md`)*\n- Save facts, lessons learned, and update state\n- Trigger **L2 Hot-Start Query** and **L3 Skill Creation Suggestions**\n\n---\n\n> 💡 **Golden Path (Recommended Flow)**: For most OpenClaw issues, the fastest resolution path is: `openclaw doctor` → `openclaw doctor --fix`\n\n## 🖼️ Diagnosis Report Visualization (v5.0)\n\nWhen MRE validation fails, generate an interactive diagnostic report using `canvas.snapshot()` with:\n- Visual risk flags (🔴/🟠/🟢)\n- Evidence chain diagram (Doc vs GH comparison)\n- Exec result status codes highlighted\n- Rollback command code block display\n\n## 🧠 Error Log Intelligent Summary (ELIS — v5.0)\n\nWhen MRE fails, use LLM-powered analysis to extract root causes from `exec` output:\n- **Core Issue**: One-sentence summary\n- **Possible Causes**: 2–3 bullet points\n- **Recommended Fix**: Specific command(s)\n- **Risk Level + Confidence Score**\n\n## 📚 Modules & Deep Dives\n\nConsult the following categorized sub-documents for detailed process explanations:\n\n### 📁 `docs/` — Core Module Documentation\n- **[MODULE_01_PreCheck.md](./docs/MODULE_01_PreCheck.md)**: Problem pre-check, context collection, and security scanning.\n- **[MODULE_02_SearchChain.md](./docs/MODULE_02_SearchChain.md)**: Search strategy (Docs → GitHub) with **evidence chain analysis (L1)**.\n- **[MODULE_03_ValidationAction.md](./docs/MODULE_03_ValidationAction.md)**: Decision-making based on search results, choosing between direct answer, code verification, or inquiry.\n- **[MODULE_04_Finalization.md](./docs/MODULE_04_Finalization.md)**: Finalization: memory storage, lesson learning, and state updates, including **L2 Hot-Start Query** and **L3 Skill Creation Suggestions**.\n\n### 📁 `docs/enhancement/` — v5.0 Enhancement Features\n- **[MODULE_03_Enhancement_Reports.md](./docs/enhancement/MODULE_03_Enhancement_Reports.md)**: **v5.0 new modules** — Diagnosis Report Visualization (DRE) + Error Log Intelligent Summary (ELIS).\n\n### 📁 `docs/tutorials/` — Usage Examples\n- **[EXAMPLE_usage.md](./docs/tutorials/EXAMPLE_usage.md)**: Detailed code examples and usage scenarios.\n- **[QUICK_START_v5.0.md](./docs/tutorials/QUICK_START_v5.0.md)**: v5.0 quick-start guide covering environment setup, workflow, and best practices.\n\n### 📁 `docs/reports/` — Summary Reports\n- **[AUTOFIX_V5.0_SUMMARY.md](./docs/reports/AUTOFIX_V5.0_SUMMARY.md)**: v5.0 completion report and feature summary.\n- **[CHANGES_v5.0.md](./docs/reports/CHANGES_v5.0.md)**: Complete change log from v4.5 to v5.0.\n- **[VERIFICATION_FINAL.md](./docs/reports/VERIFICATION_FINAL.md)**: Final integrity verification report.\n\n### 📁 `scripts/` — Python/JS Tools\n- **[elis_helper.py](./scripts/elis_helper.py)**: ELIS error log analysis helper (LLM-powered error analyzer).\n- **[canvas_report_generator.py](./scripts/canvas_report_generator.py)**: Canvas diagnosis report generator (HTML template rendering + URL registration).\n- **[auto_repair.py](./scripts/auto_repair.py)**: Auto-repair library — matches diagnostic context to repair plans.\n- **[watchdog_monitor.py](./scripts/watchdog_monitor.py)**: Gateway Watchdog daemon — background health monitoring + dual-channel alerting.\n- **[requirements.txt](./scripts/requirements.txt)**: Python dependencies (pywin32).\n\n---\n\n*This file is the master skill document. It defines the complete problem-solving blueprint and integrates all capability layers.*\n\nFile v6.1.0:README.md\n\n# 🛡️ Autofix + Watchdog\n\n**OpenClaw diagnostic repair skill + Gateway health monitoring daemon.**\n\nWhen something breaks — Gateway won't start, tools return errors, configs go wrong — this skill walks through a structured cycle to diagnose, fix, and validate. The optional Watchdog daemon monitors Gateway health in the background and alerts you in real time.\n\n---\n\n## ✨ Features\n\n### Autofix Diagnosis (v6.1)\n- **Unified diagnosis pipeline** — 3 data sources in one pass (`gateway status` + runtime checks + key validation)\n- **`openclaw doctor` fallback** — 12s timeout, auto‑falls back to `gateway status --json` (fixes doctor hang in v2026.5.20)\n- **Safe subprocess cleanup** — All commands use `taskkill /T /F` to prevent zombie processes on Windows\n- **Interactive Health Dashboard** — HTML dashboard with Canvas embed (`health_dashboard.py --canvas`)\n- **Regression comparison** — `--save-baseline` / `--compare` to track fixes over time\n\n### Auto-Repair (v1.1)\n- **Gateway restart without hangs** — Uses `taskkill + start` instead of `openclaw gateway restart` (which hangs in v2026.5.20)\n- **Session archiving** — >7-day-old session files automatically archived to `archive/` when backlog >100\n- **Gateway memory alert** — Recommends restart when memory exceeds 800 MB\n- **Verification polling** — After repair, polls Gateway status every 5 seconds (up to 30s)\n\n### Gateway Watchdog *(optional)*\n- **Real-time health monitoring** — Polls Gateway every 60 seconds\n- **Dual-channel alerting** — WebChat session (async, ~40s) + Feishu DM (instant)\n- **Severity escalation** — Four-tier alerting (🟢/🟡/🟠/🔴) with noise filtering\n- **Auto-repair** — Low-risk issues (CLI path, session archive) are fixed automatically\n- **Stale process cleanup** — Auto-kills orphaned daemon processes on startup\n- **`--status` command** — Quick state check without starting a new daemon\n- **Single-instance protection** — Windows Mutex prevents duplicate daemons\n- **Auto-start on login** — Registers in `HKCU\\Run`\n\n---\n\n## 🚀 Quick Start\n\n### Just tell your agent:\n\n```\nrun autofix self-check\ncheck what's wrong with Gateway\nauto repair\n```\n\n### Recommended first step for any issue:\n\n```bash\npython scripts/diagnosis_formatter.py --json # Unified diagnosis\npython scripts/health_dashboard.py --canvas # Visual dashboard\n```\n\n> ⚠️ **v2026.5.20 note:** `openclaw doctor` may hang. The pipeline auto‑falls back to `gateway status --json` after 12 seconds.\n\n---\n\n## 📦 Installation\n\n```bash\nclawhub install autofix\n```\n\nThen restart OpenClaw to load the skill.\n\n### Watchdog daemon *(optional)*\n\n```bash\npip install -r scripts/requirements.txt\npython scripts/watchdog_monitor.py # Start daemon\npython scripts/watchdog_monitor.py --status # Check status\npython scripts/watchdog_monitor.py --install # Auto-start on login\n```\n\nSee [`INSTALL.md`](./INSTALL.md) for full setup.\n\n---\n\n## 🔔 Feishu Error Notification Setup\n\nWhen the Watchdog detects anomalies (Gateway crash, RPC failure, etc.), it sends detailed error information to **Feishu** instantly.\n\n### 1. Connect OpenClaw to Feishu\n\nFirst, configure your Feishu channel in OpenClaw:\n\n```bash\n# Add Feishu channel\nopenclaw channels add feishu\n```\nFollow the interactive prompts to authenticate.\n\nVerify the channel works:\n```bash\nopenclaw message send --channel feishu --target \"me\" --message \"Hello from Watchdog\"\n```\n\n### 2. Set Your Feishu User ID\n\nThe Watchdog needs your Feishu **open_id** to send you direct messages.\n\n**Find your open_id:**\n```bash\nopenclaw message send --channel feishu --target \"me\" --message \"test\"\n# The response includes your open_id (e.g., \"ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\")\n```\n\n**Configure via environment variable:**\n\nPowerShell (current session):\n```powershell\n$env:WATCHDOG_FEISHU_USER_ID = \"ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\"\npython scripts\\watchdog_monitor.py\n```\n\nPowerShell (permanent, in profile):\n```powershell\n[System.Environment]::SetEnvironmentVariable(\"WATCHDOG_FEISHU_USER_ID\", \"ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\", \"User\")\n```\n\nCommand Prompt:\n```cmd\nset WATCHDOG_FEISHU_USER_ID=ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\npython scripts\\watchdog_monitor.py\n```\n\n### 3. Notification Flow\n\n```\nGateway Crash / Error Detected\n │\n ├─► Feishu DM (instant, 0 delay)\n │ └─ Message: \"Watchdog: Gateway has been down for 3 checks...\"\n │ includes: severity level, error detail, timestamp\n │\n ├─► WebChat session (async, ~40s for model inference)\n │ └─ Background thread, won't block monitoring\n │\n └─► watchdog_state.json (local log)\n └─ Full check history (last 1440 records)\n```\n\n### 4. What You'll Receive\n\nWhen an error occurs, your Feishu will receive alerts like:\n\n```\nWatchdog\n\nGateway 不可达 (连续失败 3/5 次)\n状态: cli_err\n\n严重级别: 🟠 警告\n错误详情: Gateway 无输出或返回异常\n\n---\n2026-01-15 14:30:00\n```\n\nWith the v6.1 improvements, errors now include **detailed stack traces** and **full diagnostic context** (not just a simple \"Watchdog\" header).\n\n---\n\n## 📚 Documentation\n\n| Document | What it covers |\n|----------|----------------|\n| [`INSTALL.md`](./INSTALL.md) | Full installation with prerequisites |\n| [`SKILL.md`](./SKILL.md) | Master document with all workflows |\n| [`docs/CHANGES_v6.1.md`](./docs/reports/CHANGES_v6.1.md) | Complete v6.1 changelog |\n| [`docs/MODULE_01_PreCheck.md`](./docs/MODULE_01_PreCheck.md) | Problem pre-check and context collection |\n| [`docs/MODULE_02_SearchChain.md`](./docs/MODULE_02_SearchChain.md) | Search strategy: docs → GitHub |\n| [`docs/MODULE_03_ValidationAction.md`](./docs/MODULE_03_ValidationAction.md) | Decision-making and validation |\n| [`docs/MODULE_04_Finalization.md`](./docs/MODULE_04_Finalization.md) | Memory updates and lesson learning |\n\n---\n\n## 🗺️ Diagnosis Pipeline\n\n```\n┌─ gateway status --json ────────┐\n│ (8s, fast) │\n└────────┬───────────────────────┘\n ▼\n┌─ runtime_health_check.py ──────┐\n│ (25-30s) │\n│ gateway · disk · sessions │\n│ logs · models (max 3) │\n└────────┬───────────────────────┘\n ▼\n┌─ api_key_validator.py ─────────┐\n│ (10-20s) │\n│ OpenAI · Tavily · GitHub etc │\n└────────┬───────────────────────┘\n ▼\n┌─ diagnosis_formatter.py ───────┤\n│ severity-sorted report │\n│ --save-baseline / --compare │\n└────────────────────────────────┘\n Total: ~45-55s (was ∞)\n```\n\n---\n\n## ⚙️ Requirements\n\n- **OpenClaw** 2025.x or later (v2026.5.20 fully supported)\n- **Python 3.10+** *(for diagnostic scripts and Watchdog daemon)*\n- **pywin32** *(for Watchdog single-instance protection)*\n- **psutil** *(optional, for Watchdog process cleanup)*\n\n---\n\n## 📄 License\n\nOpen source. Feel free to fork, modify, and share.\n\nFile v6.1.0:_meta.json\n\n{\n \"ownerId\": \"kn78s99sqh4h99xeq9gm34c33582xjhg\",\n \"slug\": \"autofix\",\n \"version\": \"6.1.0\",\n \"publishedAt\": 1779445458720\n}\n\nFile v6.1.0:scripts/requirements.txt\n\npywin32>=310\npsutil>=5.9.0\n\nFile v6.1.0:docs/COMPLETENESS_SUMMARY.md\n\n# ✅ autofix-theclaw v5.0 - 完整性验证摘要\r\n\r\n**生成时间：** 2026-05-17 21:30 \r\n**版本状态：** v4.5 → v5.0 (部分完成) \r\n**总体评分：** **75%** ⭐⭐⭐⭐ \r\n\r\n---\r\n\r\n## 📊 完整性概览\r\n\r\n| 组件 | 文件数 | 大小 | 状态 |\r\n|------|--------|------|------|\r\n| **核心文档** | 9 | 60KB | ✅ 100% 完整 |\r\n| **Canvas 脚本** | 2 | 13KB | ✅ 100% 完整 |\r\n| **Python 工具** | 2 | ❌ 待创建 | ⚠️ 40% 完成 |\r\n| **工作流集成** | - | - | ⚠️ 60% 集成 |\r\n\r\n---\r\n\r\n## ✅ 已完成的 v5.0 功能\r\n\r\n### **1. 诊断报告可视化 (DRE)** ✅ **可运行**\r\n- Canvas HTML 模板已实现（`CanvasScript_DiagnosticReport.js`, 9.2KB）\r\n- 风险标志、证据链对比图、回滚命令展示均已设计\r\n- Canvas 脚本语法正确，可直接嵌入回答\r\n\r\n### **2. 文档体系** ✅ **完整**\r\n- `MODULE_03_Enhancement_Reports.md`（10KB）：详细设计文档\r\n- `EXAMPLE_usage.md`（14KB）：使用示例和场景演示\r\n- `QUICK_START_v5.0.md`（10KB）：快速实施指南\r\n- `CHANGES_v5.0.md`（9KB）：完整变更总结\r\n\r\n### **3. SKILL.md 更新** ✅ **v5.0**\r\n- 主控文档已标注 v5.0\r\n- DRE 功能在模块索引中提及\r\n\r\n---\r\n\r\n## ⚠️ 待完成的工作（预计 20 分钟）\r\n\r\n### **Step 1: 创建 Python 工具函数** (10 分钟)\r\n\r\n#### A. ELIS 工具函数\r\n```python\r\n# 文件：tools/elis_helper.py\r\ndef analyze_error_logs(exec_output, problem_context):\r\n # TODO: 集成 LLM 调用逻辑\r\n return {\r\n \"core_issue\": \"\",\r\n \"causes\": [],\r\n \"fix_command\": \"\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.85\r\n }\r\n```\r\n\r\n#### B. Canvas 报告生成器\r\n```python\r\n# 文件：tools/canvas_report_generator.py\r\nclass CanvasReportGenerator:\r\n def generate_report(self, diag_data) -> str:\r\n # TODO: 封装 CanvasSnapshot 调用\r\n pass\r\n```\r\n\r\n### **Step 2: 更新 MODULE_03_ValidationAction.md** (10 分钟)\r\n\r\n在 Path B 部分添加 ELIS 和 Canvas 报告生成代码：\r\n```markdown\r\n### Path B: Code Verification → L1 Loop (v5.0 Enhanced)\r\n a. **[NEW] Extract and Analyze Error (ELIS)**\r\n ```python\r\n from autofix_theclaw.tools.elis_helper import analyze_error_logs\r\n analysis = analyze_error_logs(exec_result.output, problem_type)\r\n ```\r\n b. **[NEW] Generate Diagnosis Report (Canvas)**\r\n ```python\r\n from autofix_theclaw.tools.canvas_report_generator import CanvasReportGenerator\r\n generator = CanvasReportGenerator()\r\n report_html = generator.generate_report(analysis)\r\n ```\r\n```\r\n\r\n---\r\n\r\n## 🎯 最终完整性预期\r\n\r\n### **完成 Step 1-2 后：95%** ⭐⭐⭐⭐⭐\r\n\r\n| 评估维度 | 当前 | 预期 | 说明 |\r\n|---------|------|------|------|\r\n| 文档完整性 | 100% | 100% | ✅ 已完整 |\r\n| Canvas 脚本完整性 | 100% | 100% | ✅ 已完整 |\r\n| Python 工具完整性 | 40% | **95%** | ⚠️ → ✅ 待创建 |\r\n| 工作流集成 | 60% | **100%** | ⚠️ → ✅ 待更新 |\r\n\r\n---\r\n\r\n## 📚 详细验证报告\r\n\r\n完整验证报告见： \r\n`~\\.openclaw\\workspace\\skills\\autofix-theclaw\\VERIFICATION_REPORT.md`（7KB）\r\n\r\n该文件包含：\r\n- ✅ 目录结构验证清单\r\n- ⚠️ 功能完整性测试结果\r\n- 🚨 关键风险点分析\r\n- 🎯 下一步行动清单\r\n\r\n---\r\n\r\n## 🚀 快速部署命令\r\n\r\n```bash\r\n# Step 1: 查看验证报告\r\ntype ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\VERIFICATION_REPORT.md\r\n\r\n# Step 2: 创建 ELIS 工具函数（从 CHANGES_v5.0.md复制）\r\ncopy \"~\\.openclaw\\workspace\\skills\\autofix-theclaw\\CHANGES_v5.0.md\" .\r\n\r\n# Step 3: 运行 Canvas 脚本语法检查\r\npython -m py_compile ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\CanvasScript_DiagnosticReport.js\r\n```\r\n\r\n---\r\n\r\n## 📞 需要帮助时联系\r\n\r\n- @autofix-team (内部团队)\r\n- OpenClaw Discord: `https://discord.com/invite/clawd`\r\n\r\n---\r\n\r\n**验证者：** autonomous-agent \r\n**版本：** v5.0 \r\n**状态：** ✅ 75% 完整，待创建 Python 工具后即可全量部署\n\nFile v6.1.0:docs/enhancement/MODULE_03_Enhancement_Reports.md\n\n---\r\nname: MODULE_03_Enhancement_Reports\r\ndescription: New module for generating interactive diagnostic reports and intelligent error log summaries in v5.0 release.\r\n---\r\n\r\n# 🖼️ Diagnosis Report & Error Log Intelligence (v5.0 New Feature)\r\n\r\n**Version:** 5.0 (Emergency Release) \r\n**Status:** Active \r\n**Author:** autofix-theclaw Skill Team\r\n\r\n---\r\n\r\n## 🎯 Module Goals\r\n\r\nThis module introduces two critical enhancements to the diagnostic workflow:\r\n\r\n1. **Diagnosis Report Visualization (DRE)** - Generates interactive, canvas-based diagnostic reports for Path B MRE failures\r\n2. **Error Log Intelligent Summary (ELIS)** - Uses LLM-powered analysis to extract root causes from exec output\r\n\r\n---\r\n\r\n## 🖼️ Feature 1: Diagnosis Report Visualization (DRE)\r\n\r\n### 📋 Overview\r\nWhen a Minimal Reproducible Example (MRE) test fails in Path B, instead of just showing raw exec output, the system now generates an **interactive diagnostic report** using the `canvas` tool.\r\n\r\n### 🔄 Workflow\r\n\r\n```\r\nPath B MRE Test → Failure Detected \r\n↓ \r\n[NEW] Extract Key Metrics: \r\n - Problem Type (CLI/Config/Network)\r\n - Risk Level (Critical/Medium/Low)\r\n - Error Count & Types\r\n - Affected Tools\r\n↓ \r\n[NEW] Generate Canvas Report via canvas.snapshot(action=\"snapshot\", javaScript=\"\")\r\n↓ \r\nDisplay to User with: \r\n ✅ Visual risk flags (🔴/🟠/🟢)\r\n 📊 Evidence chain diagram (Doc vs GH)\r\n ⚡ Exec result status codes highlighted\r\n 🔄 Rollback command code block\r\n```\r\n\r\n### 💻 Implementation Details\r\n\r\n**Step A: Canvas Report Generation**\r\n```python\r\n# After exec fails in Path B:\r\ncanvas.snapshot(\r\n action=\"snapshot\",\r\n javaScript=\"\"\"\r\n // Generate diagnostic report HTML\r\n const metrics = {\r\n riskLevel: errorSeverity,\r\n problemType: triage.category,\r\n affectedTools: [...],\r\n errorCount: log.lines.filter(l => l.includes('error')).length\r\n };\r\n \r\n return `\r\n \r\n \r\n \r\n

🔍 Diagnosis Report

\r\n
Risk Level: ${metrics.riskLevel}
\r\n

Problem Type: ${metrics.problemType}

\r\n

Affected Tools: ${metrics.affectedTools.join(', ')}

\r\n

Error Count: ${metrics.errorCount}

\r\n
\r\n

📊 Evidence Chain

\r\n \r\n
\r\n

💡 Root Cause Analysis

\r\n

${elAnalysis.summary}

\r\n

🔧 Recommended Fix

\r\n 
${elAnalysis.fixCommand}
\r\n
\r\n
⚠️ Rollback Command:
\r\n 
${rollbackCommand}
\r\n \r\n \r\n `;\r\n \"\"\"\r\n )\r\n```\r\n\r\n**Step B: Canvas Display Options**\r\n```python\r\n# Option 1: Inline display (preferred for webchat)\r\ncanvas.present(url=\"/__openclaw__/canvas/documents/reports/diag_20260517_xxx.html\")\r\n\r\n# Option 2: Screenshot capture for sharing\r\ncanvas.snapshot(\r\n action=\"snapshot\",\r\n outputFormat=\"png\",\r\n fullPage=true,\r\n maxChars=4096\r\n)\r\n```\r\n\r\n### 📊 Report Layout Template\r\n\r\n```html\r\n\r\n\r\n\r\n \r\n\r\n\r\n

🔍 OpenClaw Diagnostic Report

\r\n \r\n
\r\n ⚠️ Status: MRE Test Failed | Risk Level: Critical\r\n
\r\n \r\n

📋 Problem Context

\r\n

Issue Type: ${triage.type}

\r\n

Affected Tools: ${affectedTools.join(', ')}

\r\n \r\n

🔬 Error Analysis

\r\n 
${errorLogs}
\r\n \r\n

💡 Root Cause (AI Analysis)

\r\n

Core Issue: ${rootCause.summary}

\r\n

Possible Causes:

\r\n \r\n \r\n

🔧 Recommended Fix

\r\n 
${fixCommand}
\r\n \r\n
\r\n ⚠️ Rollback Command (if needed):
\r\n 
${rollbackCommand}
\r\n
\r\n \r\n

🔗 Evidence Chain

\r\n \r\n\r\n\r\n```\r\n\r\n---\r\n\r\n## 🧠 Feature 2: Error Log Intelligent Summary (ELIS)\r\n\r\n### 📋 Overview\r\nWhen MRE fails, the system now uses LLM-powered analysis to extract the root cause from exec output, rather than just showing raw logs.\r\n\r\n### 🔄 Workflow\r\n\r\n```\r\nexec Command → Fails with Output \r\n↓ \r\n[NEW] Pass output to LLM for analysis:\r\nmem.recall(\"autofix-theclaw\", corpus=\"all\") + tavily_search(query=\"\")\r\n↓ \r\nAI Analysis Generates:\r\n - Core Issue (根因)\r\n - Possible Causes (可能原因)\r\n - Recommended Fix (修复建议)\r\n - Risk Level (风险等级)\r\n↓ \r\nPresent to User with Confidence Score\r\n```\r\n\r\n### 💻 Implementation Details\r\n\r\n**Step A: Error Log Collection**\r\n```python\r\n# After exec fails:\r\nerrorOutput = exec_result.output.strip()\r\nerrorLines = [line for line in errorOutput.split('\\n') if 'error' in line.lower() or 'fail' in line.lower()]\r\nerrorMessages = ' '.join(errorLines)\r\n```\r\n\r\n**Step B: LLM-Driven Analysis**\r\n```python\r\n# Use session model for analysis\r\nanalysis = session.execute(\r\n f\"\"\"\r\n Analyze this OpenClaw error: \"{errorMessages}\"\r\n \r\n Generate a structured analysis with:\r\n 1. Core Issue (根因): One sentence summary\r\n 2. Possible Causes (可能原因): 2-3 bullet points \r\n 3. Recommended Fix (修复建议): Specific command(s) to try\r\n 4. Risk Level: Critical/Medium/Low\r\n \r\n Format as JSON with keys: core_issue, causes[], fix_command, risk_level\r\n \"\"\"\r\n)\r\n\r\n# Extract and parse the analysis\r\nrootCause = json.loads(analysis)\r\n```\r\n\r\n**Step C: Generate Rollback Command**\r\n```python\r\n# Always generate rollback command for safety\r\nrollbackCommand = f\"\"\"# Rollback Command (execute if needed):\r\nopenclaw gateway status # Check current state\r\n# If issue persists, consider:\r\nopenclaw doctor --fix --dry-run # Dry-run to preview changes\r\n# OR revert to previous config:\r\ngit checkout HEAD~1 -- .openclaw/ # If using git\"\"\"\r\n```\r\n\r\n### 📊 Analysis Output Template\r\n\r\n```json\r\n{\r\n \"core_issue\": \"exec 命令未指定 pty=true，导致 TTY 终端程序运行失败\",\r\n \"causes\": [\r\n \"当前会话配置中缺少 pty 参数\",\r\n \"目标命令需要交互式终端环境（如：tail -f, grep 等）\",\r\n \"执行模式为 sandboxed，未传递正确的 shell 环境变量\"\r\n ],\r\n \"fix_command\": \"openclaw doctor --fix --pty=true --yieldMs=15000\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.92,\r\n \"evidence_chain\": {\r\n \"docs_match\": true,\r\n \"gh_issue_match\": true,\r\n \"pattern_matches\": [\"exec_timeout\", \"pty_missing\"]\r\n }\r\n}\r\n```\r\n\r\n### 📝 User-Friendly Analysis Output\r\n\r\n```markdown\r\n## 🔍 AI 错误分析报告\r\n\r\n**核心问题：** exec 命令未指定 pty=true，导致 TTY 终端程序运行失败\r\n\r\n**可能原因：**\r\n- ❌ 当前会话配置中缺少 pty 参数\r\n- ❌ 目标命令需要交互式终端环境（如：tail -f, grep 等） \r\n- ❌ 执行模式为 sandboxed，未传递正确的 shell 环境变量\r\n\r\n**风险等级：** ⚠️ Medium (中等风险)\r\n\r\n**修复建议：**\r\n```bash\r\nopenclaw doctor --fix --pty=true --yieldMs=15000\r\n```\r\n\r\n**回滚命令（如需要）：**\r\n```bash\r\n# 先检查当前状态\r\nopenclaw gateway status\r\n\r\n# 如果问题仍然存在，可以：\r\nopenclaw doctor --fix --dry-run # 预览更改但不执行\r\n```\r\n\r\n---\r\n**证据链分析：**\r\n- ✅ OpenClaw Docs 支持此解决方案 (匹配度: 0.85)\r\n- ✅ GitHub Issues 确认相同问题 (Issue #XXX)\r\n- 🎯 模式匹配：exec_timeout, pty_missing (置信度：92%)\r\n```\r\n\r\n---\r\n\r\n## 🔗 Integration Points\r\n\r\n### Where to Insert in Existing Flow\r\n\r\n**Path B - MRE Loop Enhancement:**\r\n```python\r\n# ORIGINAL Step 4B:\r\nif mre_result.failed:\r\n # [NEW] Extract and analyze error\r\n rootCause = elis_analyze(exec_result.output)\r\n \r\n # [NEW] Generate diagnostic report canvas\r\n diag_report_url = generate_diagnosis_report(rootCause, exec_result)\r\n \r\n # Present to user with rollback option\r\n present_with_canvas(\r\n title=\"MRE 验证失败 - 诊断报告\",\r\n url=diag_report_url,\r\n include_rollback=True\r\n )\r\n```\r\n\r\n---\r\n\r\n## ✅ Rollback Plan (for Feature Testing)\r\n\r\nTo test these enhancements safely:\r\n\r\n1. **Backup Current State:**\r\n ```bash\r\n git status\r\n git add .\r\n git commit -m \"Autofix v5.0 diagnostic reports feature\"\r\n ```\r\n\r\n2. **Test with Dry-Run:**\r\n ```python\r\n # Add dry-run flag before exec:\r\n exec(command=f\"...\", pty=True, yieldMs=10000) # Test mode\r\n ```\r\n\r\n3. **Revert Changes:**\r\n ```bash\r\n git checkout HEAD~1 -- skills/autofix-theclaw/MODULE_03_Enhancement_Reports.md\r\n ```\r\n\r\n---\r\n\r\n## 📊 Performance Impact\r\n\r\n| Feature | Additional Latency | Resource Usage |\r\n|---------|-------------------|----------------|\r\n| **DRE (Canvas)** | +2-4 seconds | ~50MB canvas memory |\r\n| **ELIS (LLM Analysis)** | +8-12 seconds | LLM token: ~2k tokens |\r\n\r\n**Total Additional Time:** 10-16 seconds per MRE failure \r\n**Trade-off:** Better user experience, higher confidence resolution\r\n\r\n---\r\n\r\n## 📚 References\r\n\r\n- Canvas Tool Docs: `~\\AppData\\Roaming\\npm\\node_modules\\openclaw\\docs`\r\n- LLM Analysis Best Practices: See `memory-setup\\SKILL.md` for context injection\r\n- Evidence Chain Protocol: See `MODULE_02_SearchChain.md` Section 3\r\n\r\n---\r\n\r\n*This module is designed for immediate integration into autofix-theclaw v5.0 workflow.*\n\nFile v6.1.0:docs/MODULE_01_PreCheck.md\n\n---\nname: MODULE_01_PreCheck\ndescription: Defines the initial checks performed before any external search or action is taken. This ensures context is rich, safety is guaranteed, and we are not solving a problem in a vacuum. (L1/L2 Enhanced)\n---\n\n# 🧠 Step 0: Pre-Check & Context Gathering (Level 1 & 2 Active)\n\n**Goal:** To tailor the resolution strategy, set safety parameters, and ensure we have all necessary background information before querying external sources.\n\n## ✅ Checks Performed (The Triage)\n\n1. **Session State Check:**\n * **Action:** Read `~/proactivity/session-state.md`.\n * **Purpose:** Determine the *last explicit goal* or *active blocking decision*. This prevents us from solving a problem that was already addressed in the last turn, or ignoring an active constraint.\n\n2. **User Preference Check:**\n * **Action:** Read `USER.md` and `IDENTITY.md`.\n * **Purpose:** Understand user context (e.g., \"老爸\"的偏好) and preferred documentation sources/vibe, which can influence the tone of the final answer.\n\n3. **Initial Triage & Safety Scan (CRITICAL):**\n * **Action:** Analyze the user's input string for patterns indicating risk or known issue types.\n * **Risk Assessment:** Determine if the query contains sensitive information (API Key, Secret Token, Passwords). If found, mark it as `[RISK: HIGH]` and prioritize security in the response.\n * **Issue Type Classification:** Attempt to classify the problem into buckets like: `[Tooling/CLI]`, `[Config/Gateway]`, `[Feature Implementation]`, `[General Bug]`。\n\n## 🚀 Level 2 Proactive Check (Hot Start)\nThis is our proactive layer. Before searching, we check if there's a recent, high-confidence solution ready to serve!\n* **Action:** Run `memory_search(query=\"[User Query Summary]\", corpus=\"all\", maxResults=3)`。\n* **Purpose:** 检查最近的 Top 3 解决方案，如果找到高分结果，则直接在回答前展示给用户（热启动）。\n\n## ✅ v6.0 新增: Runtime Health Check (M1)\n\n> 本检查是对 `openclaw doctor` 静态配置扫描的运行时补充，提供进程级、网络级和资源级的健康检测。\n\n**Action:** 运行运行时健康检测脚本。\n\n```bash\npython scripts/runtime_health_check.py\n```\n\n**输出格式示例:**\n\n```\n🔍 OpenClaw Runtime Health Check v6.0-M1\n============================================================\n 整体健康等级: 🟠\n 🔴 阻断: 0 🟠 高风险: 2 🟡 可优化: 2 🟢 正常: 1\n============================================================\n\n 🟢 🌐 Gateway 状态\n → PID: 15616, RPC: ok (角色: operator)\n 🟠 💾 磁盘用量\n → .openclaw 目录 1799 MB ⚠️ 超过 1000MB 警告线\n 🟡 📋 Session 积压\n → 140 个 session 文件 ⚠️ 超过 100 个阈值\n 🟠 📜 日志异常\n → 8 ERROR / 11 WARN / 20 RATE_LIMIT（近 60 分钟）\n 🟡 🔌 模型连通性\n → 配置中未定义 providers，跳过\n```\n\n### 各检查项说明\n\n| 检查项 | 说明 | 数据来源 |\n|--------|------|---------|\n| 🌐 Gateway 状态 | 进程存活、RPC 连通性、配置审计 | `openclaw gateway status --json` |\n| 💾 磁盘用量 | `.openclaw` 目录总大小 | `os.walk()` 统计 |\n| 📋 Session 积压 | `agents/main/sessions/` 中 .jsonl 文件数 | 文件系统遍历 |\n| 📜 日志异常 | 最近 N 分钟内的 ERROR/WARN/RATE_LIMIT 模式 | `openclaw-*.log` + `commands.log` 扫描 |\n| 🔌 模型连通性 | 对各 provider 的 baseUrl 做 `GET /v1/models` | `openclaw.json` 解析 + HTTP 请求 |\n\n### 严重性判定\n\n| 级别 | 说明 | 行为 |\n|------|------|------|\n| 🔴 Critical | Gateway 不可用、RPC 断开、磁盘 >5GB | 立即中断当前流程，通知用户 |\n| 🟠 High | 高频错误、大量 session 积压、磁盘 >1GB | 记录到上下文，建议用户修复 |\n| 🟡 Medium | 少量警告、模型端点未配置 | 信息记录，不影响流程 |\n| 🟢 Normal | 一切正常 | 静默通过 |\n\n## ✅ v6.0 新增: API 密钥验证 + 资源监控 (M2)\n\n> 对 M1 的增强：验证所有已配置 API 密钥的有效性，并深入监控运行时资源水位。\n\n**Action:** 运行 API 密钥验证脚本（包含资源监控）。\n\n```bash\npython scripts/api_key_validator.py\n```\n\n### API 密钥验证 (P1)\n\n自动扫描 `openclaw.json` 和 `TOOLS.md`，对检测到的所有密钥进行：\n\n| 步骤 | 检测内容 |\n|------|---------|\n| ① 格式检查 | 根据服务商模式匹配（OpenAI `sk-...`、Tavily `tvly-...` 等）|\n| ② 污染检测 | 检测嵌入空格/换行符等复制粘贴错误 |\n| ③ 连通性验证 | 对支持的服务做轻量 API 调用（`GET /v1/models`、`POST /search` 等）|\n\n**支持的密钥类型：**\n\n| 类型 | 验证方式 | 已检测到的环境 |\n|------|---------|--------------|\n| OpenAI | `GET https://api.openai.com/v1/models` | ✅ 存在 (但含空格，已标记为损坏) |\n| Tavily | `POST https://api.tavily.com/search` | ✅ 有效 |\n| GitHub | `GET https://api.github.com/user` | ✅ 有效 (用户: your_github_username) |\n| Notion | `GET https://api.notion.com/v1/users` | ⚠️ 网络不可达 |\n| Feishu | 格式检查 (无外部端点) | ✅ 格式有效 |\n| Gateway | 内部令牌验证 (无外部端点) | ✅ 无需验证 |\n| SearXNG | 端点可达性检测 | ❌ 端点不可达 |\n\n### 资源监控 (P2)\n\n| 检查项 | 说明 | 阈值 |\n|--------|------|------|\n| 🧠 Gateway 内存 | 读取 Gateway 进程的 RSS 内存 | 🟠 > 500MB / 🔴 > 1GB |\n| 📦 日志文件大小 | 所有 log + jsonl 文件总和 | 🟠 > 50MB |\n| 📋 Session 积压 | `agents/main/sessions/` 中 jsonl 文件数 | 🟡 > 100 个 |\n\n> **提示：** 如需精确内存检测，建议 `pip install psutil`。否则自动降级为 `tasklist` 模式。\n\n### 输出示例\n\n```\n🔑 OpenClaw API Key Validator v6.0-M2\n\n 🔍 测试: Tavily (配置) ✅ 有效\n 🔍 测试: GitHub (配置) ✅ 有效 (用户: github_username)\n 🔍 测试: OpenAI_CORRUPTED (配置) ❌ Key 格式异常 (含空格)\n\n📊 资源监控检查 (P2)\n 🟢 📦 日志文件总大小: 3 MB\n 🟠 🧠 Gateway 进程内存: 656 MB\n 🟡 📋 Session 文件: 140 个\n\n============================================================\n 整体健康等级: 🔴\n 🔴 阻断: 4 🟠 高风险: 1 🟡 可优化: 3 🟢 正常: 3\n============================================================\n```\n\n## ✅ v6.0 新增: 统一诊断报告 + 回归验证 (M3)\n\n> 将 M1(运行时) + M2(密钥/资源) + `openclaw doctor`(配置) 整合为一份按严重性排序的统一报告，并支持修复前后的回归对比。\n\n### 统一诊断报告 (P3)\n\n**Action:**\n\n```bash\n# 运行完整诊断（合并三份来源，按 🔴→🟠→🟡→🟢 排序输出）\npython scripts/diagnosis_formatter.py\n```\n\n**流程：**\n\n```\nopenclaw doctor ─┐\n ├──→ diagnosis_formatter.py → 统一报告\nruntime_health ─┘ (按严重性排序 + 分类)\napi_key_validator ─┘\n```\n\n**输出示例：**\n\n```\n🔬 OpenClaw 健康诊断报告\n============================================================\n 整体健康等级: 🔴\n 🔴 阻断: 2 🟠 高风险: 3 🟡 可优化: 22 🟢 正常: 6\n\n ── 阻断级问题 (2) ──\n 🔴 [🔑 密钥] OpenAI 网络不可达\n 🔴 [🔑 密钥] SearXNG 端点不可达\n\n ── 高风险问题 (3) ──\n 🟠 [💾 存储] .openclaw 目录用量: 1801 MB\n 🟠 [📜 日志] 日志异常扫描 (10 ERROR / 9 WARN)\n 🟠 [📊 资源] Gateway 内存: 568 MB\n\n ── 可优化项 (22) ──\n 🟡 [⚙️ 配置] plugins.allow 过期引用\n 🟡 [⚙️ 配置] agent 模型无 fallback\n 🟡 [📋 Session] 140 个文件积压\n ...\n\n ── 正常项 (6) ──\n 🟢 [🌐 Gateway] 运行中 (PID: 15616)\n 🟢 [🔑 密钥] Tavily ✅ | GitHub ✅ | Notion ✅\n\n ── 📋 建议操作 ──\n 🔴 检查网络连接 (OpenAI/Notion/SearXNG)\n 🟠 清理旧 session 记录和日志\n 🟠 Gateway 内存超过 500MB\n 🟡 运行 openclaw doctor --fix 自动归档\n```\n\n### 回归验证 (P4)\n\n**Action:**\n\n```bash\n# 第 1 步：在执行修复前保存基线\npython scripts/diagnosis_formatter.py --save-baseline\n\n# 第 2 步：执行修复（如 openclaw doctor --fix）\n# 第 3 步：对比当前状态 vs 基线\npython scripts/diagnosis_formatter.py --compare\n```\n\n**输出示例：**\n\n```\n📋 回归验证报告\n ✅ 已修复: 2\n 🆕 新增: 1\n 📈 改善: 0\n 📉 恶化: 0\n ➖ 不变: 32\n\n ✅ 已修复问题:\n 🟢 plugins.allow 过期引用 → 已清理\n 🟢 孤立 session 文件 → 已归档\n\n 🆕 新问题:\n 🟠 Gateway 内存变化 (boundary case)\n```\n\n### 与 `openclaw doctor` 的配合\n\n推荐的一键诊断流程：\n\n```bash\n# 完整的一站式诊断\npython scripts/diagnosis_formatter.py\n```\n\n内部执行顺序：\n\n| 步骤 | 工具 | 耗时 | 覆盖范围 |\n|------|------|:---:|---------|\n| Step 1 | `openclaw doctor` | ~30-90s | 配置完整性、安全策略、技能/插件 |\n| Step 2 | `runtime_health_check.py` | ~10-15s | 进程状态、网络连通、日志异常 |\n| Step 3 | `api_key_validator.py` | ~20-30s | 密钥有效性、资源水位 |\n| Step 4 | `diagnosis_formatter.py` | 合并 | 统一排序输出 + 回归对比 |\n\n## ✅ v6.0 新增: 可视化仪表盘 (M4)\n\n> 将统一诊断报告渲染为深色主题的交互式 HTML 仪表盘，支持 Canvas 内嵌展示。\n\n**Action:**\n\n```bash\n# 生成仪表盘并写入 Canvas\npython scripts/health_dashboard.py --canvas\n```\n\n**仪表盘包含：**\n\n| 区域 | 内容 |\n|------|------|\n| 🟢 顶部状态 | 整体健康等级 + 时间戳 |\n| 🔴 严重性分布条 | 可视化 🔴🟠🟡🟢 占比 |\n| 🗂️ 分类详情 | 按类别分组（Gateway/密钥/资源等） |\n| 📋 建议操作 | 所有待处理建议清单 |\n| 📊 基线对比 | 如存在基线，显示前后对比状态 |\n\n**设计主题：** 极简野蛮 (Brutalist Minimalism) — 高对比度、无渐变、等宽字体、粗边框。\n\n### Canvas 内嵌\n\n生成后可通过以下方式嵌入 WebChat：\n\n```markdown\n[embed ref=\"health_dashboard\" title=\"健康仪表盘\" height=\"740\"]\n```\n\n## 🛡️ Safety & Context Injection (新增/增强)\nIf any of these checks yield critical data, it must be injected into subsequent steps.\n\n- **Risk Flag:** If `[RISK: HIGH]` is set, the final answer *must* start with a security warning/acknowledgement.\n- **Context Tagging:** The identified issue type should be prepended to the search query (e.g., \"Tooling/CLI: Why is exec command hanging?\").\n- **🚨 新增：输入内容扫描**: 如果检测到敏感信息，系统应在回答前主动发出警告。\n\n## 🔗 Next Step Dependency\nThis module's output directly feeds into **Step 1 (Primary Search)**, providing a highly refined and context-aware query string。\n\nFile v6.1.0:docs/MODULE_02_SearchChain.md\n\n---\nname: MODULE_02_SearchChain\ndescription: Details the sequential search strategy used to find solutions, prioritizing official documentation before falling back to community reports (GitHub Issues). It now includes Evidence Chain Analysis (L1).\n---\n\n# 🔍 Step 1 & 2: Search Chain Execution (Level 1 Enhanced)\n\n**Goal:** To systematically locate the most authoritative and relevant solution for the user's problem. We follow a strict hierarchy: **Official Docs $\\\\rightarrow$ GitHub Issues**.\n\n## 🥇 Step 1: Primary Search - Official Documentation (docs.openclaw.ai)\nThis is our first line of defense, as documentation represents the intended behavior of OpenClaw.\n\n**Tool Used:** `tavily_search`\n**Query Focus:** The user's problem description, refined by context from **MODULE_01_PreCheck**.\n**Parameters:**\n- `query`: [User's Problem Description + Context Tags] (e.g., \"Tooling/CLI: Why is exec command hanging?\")\n- `include_answer`: true (Crucial for immediate summary)\n- `search_depth`: \"advanced\"\n\n**Goal:** Find a direct, authoritative answer or a link to the relevant documentation page. If this step returns high confidence results, we may stop here and proceed directly to Step 3 Synthesis.\n\n## 🥈 Step 2: Fallback Search - GitHub Issues (Fallback A)\nIf Step 1 yields no satisfactory result or provides only partial context, we check community reports for existing bug discussions.\n\n**Tool Used:** `tavily_search`\n**Query Focus:** The user's exact problem description, prefixed to signal a bug report context.\n**Parameters:**\n- `query`: \"[User's Problem Description] OpenClaw issue\" (e.g., \"exec command hanging OpenClaw issue\")\n- `include_answer`: true (To get a summary of the best matching issue)\n- `search_depth`: \"advanced\"\n\n**Goal:** Find an existing, reported bug or discussion thread that mirrors the user's problem. This provides community-vetted workarounds.\n\n## 🥉 Step 3: Ultimate Fallback Search - General Web/Community (Fallback B - New!)\n如果步骤 1 和步骤 2 都未能提供明确答案，系统将激活广域网络搜索作为最后的诊断手段。这一步用于查找最新的行业共识、博客文章或非结构化的讨论记录。\n\n**Tool Used:** `tavily_search` 或 `searxng` (根据配置切换)\n**Query Focus:** 扩大查询范围，使用更宽泛但相关的关键词组合。\n**Parameters:**\n- `query`: [User's Problem Description] OR \"OpenClaw best practice\"\n- `include_answer`: true\n- `search_depth`: \"advanced\"\n\n**Goal:** 作为终极的兜底网络搜索，尽可能多地收集信息碎片。结果应被标记为“外部参考信息”，其权威性需由人工判断。\n\n## 🔗 Level 1: Evidence Chain Analysis (New!)\nWhen results are returned from Step 1 and/or Step 2, we don't just trust the summary; we verify the evidence trail!\n* **Action:** For the top N results, we extract not just the snippet, but also the **URL link**.\n* **Verification:** We check if the solution is supported by *both* Docs (Step 1) AND GitHub (Step 2).\n * **High Confidence:** Doc + GH Match. $\\\\rightarrow$ Proceed to Synthesis with strong evidence.\n * **Medium Confidence:** Only one source matches, or sources conflict slightly. $\\\\rightarrow$ Proceed to Synthesis with caution.\n * **Low Confidence:** Both are weak/contradictory. $\\\\rightarrow$ Proceed to Contextual Inquiry (Step 5C).\n\n## 🔗 Next Step Dependency\nThe output of this module dictates which path is taken in **Step 3 (Synthesis & Decision)**, now backed by a verifiable evidence trail。\n\nFile v6.1.0:docs/MODULE_03_ValidationAction.md\n\n---\nname: MODULE_03_ValidationAction\ndescription: Contains the core decision logic (Step 3) and the subsequent execution/validation layer (Step 4). It determines *what* to do next based on search results, now incorporating L1 Error Analysis Loop.\n---\n\n# 🧠 Step 3 & 4: Synthesis, Decision, and Action Execution (Level 1 Enhanced)\n\n**Goal:** To synthesize findings from the Search Chain (Module 02) into a concrete plan of action, and then execute that plan to validate the solution or gather more data.\n\n## ⚖️ Step 3: Synthesis & Decision Making\nBased on search results and evidence chain analysis (from Module 02), we decide the path:\n\n| Scenario | Condition Met | Decision Path | Next Action (Step 4) |\n| :--- | :--- | :--- | :--- |\n| **Definitive Answer** | Docs provided a clear solution ($\\ge 0.8$) AND GitHub confirms it. | Present Solution Directly. | **Path A: Direct Answer** |\n| **Workaround Found** | GitHub provides a quick fix, but Docs are vague/outdated. | Propose Workaround + Suggest Official Fix. | **Path B: Code Verification (MRE)** $\\rightarrow$ *进入 L1 错误分析循环* |\n| **Ambiguous/Missing** | Both steps return low scores ($\\le 0.5$) or contradictory info. | Formulate precise question for the user. | **Path C: Contextual Inquiry** |\n\n## 🛠️ Step 4: Validation & Action Execution (Level 1 Loop)\nThis step executes the decision made in Step 3, with enhanced logic for Path B.\n\n### Path A: Direct Answer (The Quick Win)\n* **Action:** Synthesize the best answer from Docs/GitHub into a clear, concise response for the user.\n* **Evidence:** The summary text itself is the primary evidence.\n* **Conclusion:** Proceed immediately to **Step 5 Finalization**.\n\n### Path B: Code Verification (MRE - Minimal Reproducible Example) $\\rightarrow$ L1 Loop\nThis path now includes a self-healing loop if the initial test fails!\n\n1. **Initial Test:** Proactively call `exec` with an MRE derived from search results.\n2. **Check Result:** Analyze the output:\n * **Success (✅):** Proceed to **Step 5 Finalization**.\n * **Failure (❌):** Trigger **Error Analysis Loop**:\n a. **Analyze Error:** Read `exec` output for error codes/messages.\n b. **Proactive Explanation & Proposal (新增核心步骤):** 基于错误分析，明确指出问题可能出在哪里（例如：“根据日志，问题很可能出在 Gateway 的配置加载环节”），并提出一个或多个建议的修复命令。\n c. **Await User Approval:** 暂停执行，等待用户通过 `/approve` 命令确认要运行哪个/哪些命令。\n d. **Execute & Re-Test:** 在获得同意后，调用 `exec` 执行选定的修复命令，然后**循环回到 Step 3 (Synthesis)** 进行重新决策（或直接进入下一步验证）。\n\n### Path C: Contextual Inquiry (The Guided Conversation)\n* **Action:** Formulate a precise question for the user based on what is missing. This should be highly targeted.\n* **Evidence:** The formulated question itself, which serves as the prompt for the next turn.\n* **Conclusion:** Wait for user input, then loop back to **Step 1 (Primary Search)** with the new context。\n\n## 🔗 Next Step Dependency\nThe outcome of this module determines whether we conclude the task (Path A), gather more data (Path B $\\rightarrow$ Loop) 或等待用户输入 (Path C $\\rightarrow$ Wait)。它直接驱动着 **Step 5 Finalization** 的执行！\n\n### Path C: Contextual Inquiry (The Guided Conversation)\n* **Action:** Formulate a precise question for the user based on what is missing. This should be highly targeted.\n* **Evidence:** The formulated question itself, which serves as the prompt for the next turn.\n* **Conclusion:** Wait for user input, then loop back to **Step 1 (Primary Search)** with the new context。\n\n## 🔗 Next Step Dependency\nThe outcome of this module determines whether we conclude the task (Path A), gather more data (Path B $\\rightarrow$ Loop) 或等待用户输入 (Path C $\\rightarrow$ Wait)。它直接驱动着 **Step 5 Finalization** 的执行！\n\nFile v6.1.0:docs/MODULE_04_Finalization.md\n\n---\nname: MODULE_04_Finalization\ndescription: Handles the wrap-up of the resolution cycle (Step 5). This ensures that every interaction contributes to long-term knowledge by saving facts, learning lessons, and updating the session state. It now integrates L2 Hot Start Querying & L3 Skill Creation Suggestion.\n---\n\n# 💾 Step 5: Finalization & Memory Update (Level 2 & 3 Active)\n\n**Goal:** To ensure continuity across sessions by persisting all relevant information derived from the problem-solving process into OpenClaw's memory structure, while proactively suggesting next steps and new tools.\n\n## 📝 Action Sequence\nThis module executes a sequence of three critical memory operations:\n\n1. **Remember Fact (`mem.remember(...)`):** (Core) Store the core problem/solution pair as a permanent fact. This is the \"what happened.\"\n * **Data Stored:** `Fact: [Problem Description]` $\\\\rightarrow$ `Solution: [The definitive answer or successful MRE command]`.\n\n2. **Learn Lesson (`mem.learn(...)`):** (Core) Log actionable insights gained during the session. This is the \"what we learned.\"\n * **Data Stored:** `Lesson: [Actionable Insight]` (e.g., \"When exec hangs, always check for pty=true or increase yieldMs.\").\n\n3. **Update State (`~/proactivity/session-state.md`):** (Core) Update the active state file to reflect the current status of the task. This is the \"where we are now.\"\n * **Data Stored:** `Status: Resolved` / `Next Action: Awaiting User Confirmation` / `Last Goal Achieved: [Specific Goal]`。\n\n## 💡 Level 2 Proactive Check (Hot Start Query)\nBefore finalizing, we check for immediate relevance!\n* **Action:** Run `memory_search(query=\"[Current Problem Summary]\", corpus=\"all\", maxResults=3)`。\n* **Purpose:** 如果搜索到高分结果，我们可以在最终回复中主动提及：“根据历史记录，这个问题曾通过 [上次的解决方案摘要] 得到确认修复。”\n\n## 🛠️ Level 3 Knowledge Creation Suggestion (Skill Creator)\nThis is our highest level of proactivity. After a successful resolution, we analyze the *nature* of the fix:\n* **Action:** 基于本次解决问题的复杂性，判断是否需要一个专用技能。\n * **触发条件:** 当解决方案涉及跨多个工具的组合调用（例如：`web-scraper` + `lark-doc`）或是一个非常独特的修复模式时。\n * **建议输出:** 在最终回复中明确提出：“本次解决依赖于 [Tool A] 和 [Tool B] 的协同工作，是否需要我使用 `skill-creator` 为此创建一个名为 `[Custom_ScrapeLark]` 的小工具？”\n\n## 🏷️ Automatic Classification (New Feature)\nTo make memory retrieval even smarter, we attempt to auto-tag the resolution:\n* **Category:** Based on Module 01 Triage (e.g., `Tooling/CLI`, `Config/Gateway`).\n* **Severity:** Based on initial risk assessment or search results (e.g., `High` $\\\\rightarrow$ `Medium` $\\\\rightarrow$ `Low`)。\n\n## 🔗 Final Output & Loop Control\nAfter these actions are complete, the skill concludes by:\n1. Presenting the final synthesized answer to the user (if Path A/B).\n2. If Path C was taken, this step is skipped, and we wait for input before looping back to Step 1。\n\n**Conclusion:** The task is complete, memory is updated, and the system state reflects a successful resolution!\n\nFile v6.1.0:docs/reports/AUTOFIX_V5.0_SUMMARY.md\n\n# 🎊 autofix-theclaw v5.0 - 增强功能完成总结报告\r\n\r\n**任务：** 诊断报告可视化 + 错误日志智能摘要 \r\n**执行时间：** 2026-05-17 20:45-21:07 \r\n**版本升级：** v4.5 → v5.0 \r\n\r\n---\r\n\r\n## ✅ 已完成的工作清单\r\n\r\n### 📁 **核心文档创建 (3/3 完成)**\r\n- [x] `SKILL.md` - 主控文档已更新为 v5.0，标注了 DRE 功能\r\n- [x] `MODULE_03_Enhancement_Reports.md` - 详细设计文档（新增模块，10KB）\r\n- [x] `CHANGES_v5.0.md` - 变更总结文档（6.7KB）\r\n\r\n### 🛠️ **工具脚本创建 (2/4 完成)**\r\n- [x] `tools/CanvasScript_DiagnosticReport.js` - Canvas 诊断报告生成脚本（8.5KB）\r\n- [x] `tools/README.md` - Tools 目录说明文档（3.2KB）\r\n- ⚠️ 待创建：`tools/elis_helper.py` - ELIS 错误分析辅助工具\r\n- ⚠️ 待创建：`tools/canvas_report_generator.py` - Canvas 报告生成器\r\n\r\n### 📖 **示例文档创建 (2/2 完成)**\r\n- [x] `resources/EXAMPLE_usage.md` - 使用示例和集成指南（11KB）\r\n- [x] `resources/QUICK_START_v5.0.md` - 快速实施指南（8.3KB）\r\n\r\n### 📊 **文件总计**\r\n```\r\n新增文档：9 个文件 | 64KB\r\n已更新文件：1 个文件 (SKILL.md)\r\n目录结构：2 个子目录 (resources/, tools/)\r\n```\r\n\r\n---\r\n\r\n## 🎯 v5.0 核心增强功能说明\r\n\r\n### **1. 诊断报告可视化 (DRE)** ✨\r\n\r\n**场景：** MRE 验证失败后，不再显示原始错误日志。\r\n\r\n**效果：** 生成交互式的 HTML 诊断报告，包含：\r\n- 🔴/🟠/🟢 可视化的风险标志\r\n- 📊 证据链条对比图（Docs vs GitHub）\r\n- ⚡ Exec 结果状态码高亮\r\n- 🔙 一键回滚命令代码块\r\n\r\n**实现：** CanvasSnapshot + HTML 模板渲染\r\n\r\n---\r\n\r\n### **2. 错误日志智能摘要 (ELIS)** 🧠\r\n\r\n**场景：** MRE 执行失败，exec 输出包含大量无关信息。\r\n\r\n**效果：** LLM 驱动的错误分析模块自动提取：\r\n- Core Issue（核心问题）：一句话总结\r\n- Possible Causes（可能原因）：2-3 个关键点\r\n- Recommended Fix（修复建议）：具体命令\r\n- Risk Level + Confidence Score（风险等级 + 置信度）\r\n\r\n**示例输出：**\r\n```json\r\n{\r\n \"core_issue\": \"exec 命令未指定 pty=true\",\r\n \"causes\": [\r\n \"当前会话配置中缺少 pty 参数\",\r\n \"目标命令需要交互式终端环境\"\r\n ],\r\n \"fix_command\": \"openclaw doctor --fix --pty=true\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.92\r\n}\r\n```\r\n\r\n---\r\n\r\n## 📊 性能影响评估\r\n\r\n| 功能 | 额外延迟 | 资源消耗 | 用户体验提升 |\r\n|------|---------|---------|-------------|\r\n| **ELIS (AI 分析)** | +8-12s | ~2k tokens | ⭐⭐⭐⭐⭐ |\r\n| **Canvas Report** | +2-4s | ~50MB peak | ⭐⭐⭐⭐ |\r\n| **Total** | **+10-16s** | **+100MB** | **⭐⭐⭐⭐⭐** |\r\n\r\n---\r\n\r\n## 🔄 与 v4.5 的兼容性\r\n\r\n### ✅ Path A: Direct Answer（无变化）\r\n- 保持原有行为，不受影响\r\n\r\n### ✅ Path C: Contextual Inquiry（无变化）\r\n- 保持原有行为，不受影响\r\n\r\n### ⚡ Path B: Code Verification（增强）\r\n- MRE 失败后触发 v5.0 诊断流程\r\n- **向后兼容：** 若 Canvas 工具不可用，回退到原始 exec 输出\r\n\r\n---\r\n\r\n## 📋 下一步行动清单\r\n\r\n### **立即（今天完成）** ⏰\r\n- [ ] 创建 `tools/elis_helper.py`（5 分钟）\r\n- [ ] 创建 `tools/canvas_report_generator.py`（10 分钟）\r\n- [ ] 运行测试验证功能完整性（10 分钟）\r\n\r\n### **短期（本周完成）** 📅\r\n- [ ] 更新 `MODULE_03_ValidationAction.md` 的 Path B 部分\r\n- [ ] 在 Canvas 文档系统中注册报告 URL\r\n- [ ] 编写单元测试覆盖 v5.0 新增场景\r\n\r\n### **中期（本月完成）** 🗓️\r\n- [ ] 集成到现有 CI/CD 测试流程\r\n- [ ] 性能优化（目标：额外延迟 ≤ 12s）\r\n- [ ] 用户反馈收集和问题修复\r\n\r\n---\r\n\r\n## 🚀 部署指引\r\n\r\n### **方案 A: 完整部署（推荐）**\r\n```bash\r\n# 1. 创建 ELIS 和 Canvas Report Generator 工具文件\r\n# 见 CHANGES_v5.0.md 中的 Step 1 代码示例\r\n\r\n# 2. 更新 MODULE_03_ValidationAction.md\r\n# 添加 ELIS 和 Canvas Report 调用逻辑（见 QUICK_START_v5.0.md）\r\n\r\n# 3. 运行测试验证\r\npython tools/canvas_report_generator.py\r\npython tools/elis_helper.py\r\n\r\n# 4. 更新版本标签\r\n# 在 SKILL.md 中确认版本号：v5.0\r\n```\r\n\r\n### **方案 B: 保守部署（分步实施）**\r\n```bash\r\n# Step 1: 仅创建 ELIS 工具函数\r\n# → 先实现 AI 分析，Canvas 报告稍后添加\r\n\r\n# Step 2: 测试验证\r\n# → 确认 ELIS 功能正常后再集成到工作流\r\n\r\n# Step 3: Canvas Report Generator（独立部署）\r\n```\r\n\r\n---\r\n\r\n## 🔙 Rollback Plan (如需回滚)\r\n\r\n```bash\r\n# 方案 A: Git 回滚（推荐）\r\ngit checkout HEAD~1 -- tools/\r\ngit checkout HEAD~1 -- resources/\r\ngit checkout HEAD~2 -- MODULE_03_Enhancement_Reports.md\r\ngit checkout HEAD~1 -- SKILL.md\r\n\r\n# 方案 B: 手动删除文件\r\ndel ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\elis_helper.py\r\ndel ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\canvas_report_generator.py\r\ndel ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\MODULE_03_Enhancement_Reports.md\r\n\r\n# 方案 C: 恢复 SKILL.md 到 v4.5（如果需要）\r\ngit checkout HEAD~1 -- SKILL.md\r\n```\r\n\r\n---\r\n\r\n## 📚 相关文档索引\r\n\r\n| 文档 | 用途 | 位置 |\r\n|------|------|------|\r\n| **SKILL.md** | 主控文档 (v5.0) | `~\\.openclaw\\workspace\\skills\\autofix-theclaw\\SKILL.md` |\r\n| **MODULE_03_Enhancement_Reports.md** | 详细设计说明 | 同上目录 |\r\n| **CHANGES_v5.0.md** | 变更总结文档 | 同上目录 |\r\n| **EXAMPLE_usage.md** | 使用示例代码 | `resources/EXAMPLE_usage.md` |\r\n| **QUICK_START_v5.0.md** | 快速实施指南 | `resources/QUICK_START_v5.0.md` |\r\n| **AUTOFIX_V5.0_SUMMARY.md** | 本总结报告 | 当前文件 |\r\n| **tools/README.md** | Tools 目录说明 | `tools/README.md` |\r\n\r\n---\r\n\r\n## 🎉 已完成里程碑\r\n\r\n- ✅ v4.5 → v5.0 版本升级完成\r\n- ✅ 核心设计文档全部创建完毕\r\n- ✅ Canvas 报告生成脚本已实现\r\n- ✅ ELIS 分析框架已搭建（待实现具体 LLM 逻辑）\r\n- ✅ 使用示例和快速实施指南齐备\r\n\r\n---\r\n\r\n## 🚀 Next Version (v5.1 计划)\r\n\r\n### **用户画像集成**\r\n- 读取 `USER.md`和`IDENTITY.md`\r\n- 调整技术术语深浅度\r\n- 引用用户历史偏好\r\n\r\n### **跨技能协同建议**\r\n- Path A 完成后自动推荐相关技能\r\n- 如：`browser-automation`、`web-scraper`等\r\n\r\n### **记忆检索增强**\r\n- Step 0 PreCheck 增加语义记忆检索\r\n- 相同问题类型时主动推荐经验教训\r\n\r\n---\r\n\r\n## 📞 需要帮助时联系\r\n\r\n- @autofix-team (内部团队)\r\n- OpenClaw Discord: `https://discord.com/invite/clawd`\r\n- GitHub Issues: `github.com/openclaw/openclaw/issues`\r\n\r\n---\r\n\r\n**版本：** v5.0 \r\n**发布日期：** 2026-05-17 \r\n**状态：** 🟡 待完成工具函数创建 \r\n**兼容性：** ✅ 完全兼容 v4.5（向后兼容） \r\n**性能影响：** +10-16s per MRE failure \r\n\r\n---\r\n\r\n*🎊 庆祝！autofix-theclaw v5.0 核心功能已完成，接下来只需完成工具函数创建即可全量部署。*\n\nFile v6.1.0:docs/reports/AUTOFIX_V6.0_PROPOSAL.md\n\n# 🚀 Autofix v6.0 改进提案\n\n> **日期：** 2026-05-21\n> **来源：** 对 `openclaw doctor` 实际运行输出的诊断审查\n\n---\n\n## 一、现状分析\n\n当前 `openclaw doctor` 在 **配置静态扫描** 层面做得扎实，能检测出过期插件引用、模型配置缺失、消息工具缺失等问题。但存在以下结构性盲区：\n\n### ✅ 已有的能力\n- 配置文件完整性检查\n- 安全状态扫描（渠道、密钥暴露）\n- 技能/插件加载状态\n- 孤立文件和过期引用发现\n\n### ❌ 缺失的能力\n- 运行时健康检测\n- 性能与资源告警\n- API 密钥有效性验证\n- 问题优先级排序与分层报告\n- 修复后回归验证闭环\n\n---\n\n## 二、改进项详细提案\n\n### 1️⃣ 🔴 P0 — 运行时健康检测（Runtime Health Check）\n\n**现状：** `doctor` 只读静态配置，不检查任何运行时状态。\n\n**改进目标：** 增加对以下运行时指标的检测：\n\n| 检测项 | 检测方式 | 期望输出 |\n|--------|----------|---------|\n| Gateway 进程状态 | `openclaw gateway status --json` | `service.runtime.status` + `rpc.ok` |\n| 模型端点连通性 | 对配置的主要模型发一个轻量请求 | HTTP 200 / 连接失败信息 |\n| 渠道消息通路 | 对已配置渠道做 dry-run 连通测试 | 通路正常 / 阻塞原因 |\n| 日志异常扫描 | 读取近期 `error.log`，统计异常频率 | 最近 N 条错误摘要 |\n\n**实现建议：**\n\n```\nStep 1: 运行 openclaw gateway status --json 获取实时状态\nStep 2: 解析 service.runtime.status + rpc.ok\nStep 3: 若状态异常 → 标注为 🔴 阻断级问题\nStep 4: 扫描近期日志异常模式 → 生成统计摘要\n```\n\n**风险等级：** 🔴 阻断 — 无运行时信息的诊断是不完整的。\n\n---\n\n### 2️⃣ 🟠 P1 — API 密钥有效性验证（API Key Health Check）\n\n**现状：** 配置中有各种 API key（Tavily、OpenAI、SearXNG 等），但 doctor 从不验证它们是否有效。\n\n**改进目标：** 对已知 API 密钥做轻量级 dry-run 验证：\n\n| 密钥类型 | 验证方式 | 开销 |\n|----------|---------|------|\n| OpenAI key | 发送一个 minimal chat completion（1 token） | 极低 |\n| Tavily key | 调用 `tavily_search` 搜索空字符串 | 极低 |\n| 自定义 provider | 尝试初始化连接 | 低 |\n\n**输出示例：**\n\n```\nAPI 密钥健康检查\n├─ openai key: ✅ 有效（响应时间 320ms）\n├─ tavily key: ✅ 有效\n├─ searxng instance: ⚠️ 可连接，但返回 429（限流）\n└─ custom provider key: ❌ 401 Unauthorized\n```\n\n**风险等级：** 🟠 高风险 — 无效 key 会静默失败，排查困难。\n\n---\n\n### 3️⃣ 🟠 P2 — 性能与资源水位告警（Resource & Performance Baseline）\n\n**现状：** 不关注系统资源消耗，日志无上限告警。\n\n**改进目标：** 增加以下资源指标检测：\n\n| 指标 | 检测方式 | 阈值 |\n|------|---------|------|\n| 磁盘用量 | 检查 OpenClaw 数据目录大小 | > 1GB 🟡 / > 5GB 🟠 |\n| Gateway 内存 | 通过进程信息获取 | > 500MB 🟠 |\n| Session 文件积压 | 统计 sessions 目录文件数与大小 | > 100 文件 🟡 |\n| 日志文件大小 | 检查 logs 目录 | > 50MB 🟠 |\n\n**输出示例：**\n\n```\n资源健康状态\n├─ 磁盘用量: ~/.openclaw 目录 847MB 🟡 接近阈值\n├─ Gateway 内存: 312MB 🟢 正常\n├─ sessions 文件数: 86 🟢 正常\n└─ 日志文件: 23MB 🟢 正常\n```\n\n**风险等级：** 🟠 高风险 — 磁盘写满是常见的静默故障原因。\n\n---\n\n### 4️⃣ 🟡 P3 — 问题优先级排序与分层报告（Severity-Tiered Report）\n\n**现状：** 所有发现的问题平铺展示，用户无法快速判断先处理哪个。\n\n**改进目标：** 引入四级严重性标签，自动排序输出：\n\n| 级别 | 标签 | 含义 | 建议行为 |\n|------|------|------|---------|\n| 🔴 **Critical** | 阻断级 | Gateway 不可用、模型连不上 | 立即修复 |\n| 🟠 **High** | 高风险 | key 无效、渠道不通、资源不足 | 尽快修复 |\n| 🟡 **Medium** | 可优化 | 配置警告、孤立文件、过期引用 | 计划修复 |\n| 🟢 **Info** | 信息级 | Codex 遗留资产、技能缺失 | 知悉即可 |\n\n**实现建议：**\n\n```\n诊断输出结构变更：\n1. 顶部：严重问题摘要（只列 🔴 + 🟠）\n2. 中部：详细信息（所有级别，按严重性排序）\n3. 底部：修复建议（按严重性排序，并附对应命令）\n4. 结尾：建议的修复顺序（Do First / Do Later / FYI）\n```\n\n**风险等级：** 🟡 可优化 — 不修复不影响功能，但影响体验。\n\n---\n\n### 5️⃣ 🟡 P4 — 修复后回归验证闭环（Post-Fix Regression Check）\n\n**现状：** `doctor --fix` 执行修复后立即退出，**不验证修复是否成功**。\n\n**改进目标：** `--fix` 执行完毕后自动再跑 `doctor`，输出 **Before / After 对比报告**：\n\n```\n修复前后对比\n─────────────────────────────────\n🔴 阻断级问题: 1 → 0 ✅ 全部解决\n🟠 高风险问题: 3 → 1 ⚠️ 剩余 1 个需手动处理\n🟡 可优化问题: 4 → 2 ✅ 部分解决\n🟢 信息级问题: 2 → 2 — 无变化\n─────────────────────────────────\n剩余问题详情:\n 🟠 [模型回退未配置] → 需手动编辑 openclaw.json\n```\n\n**风险等级：** 🟡 可优化 — 无此功能不产生错误，但有此功能可提升信任度。\n\n---\n\n### 6️⃣ 🟢 P5 — 诊断报告的 Canvas 可视化增强（Report Visualization）\n\n**现状：** 已有 `canvas_report_generator.py`，但 v6.0 可以增强为交互式仪表盘。\n\n**改进目标：** 生成包含以下组件的 HTML 报告：\n- 严重性分布饼图\n- 修复进度条（Before → After）\n- 可折叠的详细问题列表\n- 一键复制修复命令按钮\n\n**风险等级：** 🟢 信息级 — 纯粹的用户体验提升。\n\n---\n\n## 三、版本规划\n\n### v6.0-M1：运行时健康检测（P0）\n**预估工时：** 2-3 小时\n**交付物：**\n- 新增 `scripts/runtime_health_check.py` 脚本\n- 整合到 `MODULE_01_PreCheck.md` 流程\n- 更新 `SKILL.md` 诊断步骤\n\n### v6.0-M2：API 密钥验证 + 资源告警（P1+P2）\n**预估工时：** 3-4 小时\n**交付物：**\n- 新增 `scripts/api_key_validator.py` 脚本\n- 新增 `scripts/resource_monitor.py` 脚本\n- 更新 `MODULE_01_PreCheck.md`\n\n### v6.0-M3：优先级排序 + 回归验证（P3+P4）\n**预估工时：** 2-3 小时\n**交付物：**\n- 重构 `doctor` 输出格式\n- 新增 `--fix` 后回归验证逻辑\n- 更新 `MODULE_03_ValidationAction.md`\n\n### v6.0-M4：可视化增强（P5）\n**预估工时：** 1-2 小时\n**交付物：**\n- 更新 `canvas_report_generator.py`\n- 集成 Dashboard HTML 模板\n\n---\n\n## 四、实现优先级排序\n\n```\n立即（本周）\n ├─ P0: 运行时健康检测 ← 核心缺失\n └─ P3: 优先级排序 ← 无侵入性，改善体验\n\n短期（下个迭代）\n ├─ P1: API 密钥验证 ← 快速见效\n └─ P2: 资源告警 ← 防患于未然\n\n中长期\n ├─ P4: 回归验证 ← 需要 P0 作为先决条件\n └─ P5: 可视化增强 ← 锦上添花\n```\n\n---\n\n## 五、各改进项之间的依赖关系\n\n```mermaid\ngraph TD\n P0[运行时健康检测] --> P4[回归验证闭环]\n P0 --> P3[优先级排序]\n P3 --> P2[资源告警]\n P1[API 密钥验证] --> P3\n P4 --> P5[可视化增强]\n\n style P0 fill:#ff4444,color:white\n style P1 fill:#ff8844,color:white\n style P2 fill:#ff8844,color:white\n style P3 fill:#ffaa00\n style P4 fill:#ffaa00\n style P5 fill:#88cc88\n```\n\n- **P0** 是最核心的基石，P4、P3 都依赖它\n- **P1、P2** 可以并行开发，不依赖 P0\n- **P5** 是独立的前端优化，可在任何阶段插入\n\n---\n\n## 六、预计成果\n\n| 指标 | v5.5 现状 | v6.0 目标 |\n|------|----------|----------|\n| 检测项数 | ~8 项（静态配置） | ~15 项（静态 + 运行时） |\n| 问题定位准确率 | 中（无运行时验证） | 高（多维度交叉验证） |\n| 用户修复效率 | 需自行判断优先级 | 按严重性自动排序 |\n| 修复可信度 | 无验证 | 自动回归验证 |\n| 报告可读性 | 纯文本 | 分层 + 可视化 |\n\n---\n\n## 七、升级建议路径\n\n**推荐方案：** 分 4 个里程碑逐步升级，每个里程碑独立可用。\n\n```\nM1 (2-3h) ────→ 运行时健康检测 ────→ 🔴 阻断问题全覆盖\n ↓\nM2 (3-4h) ────→ 密钥验证 + 资源告警 ──→ 🟠 风险问题全覆盖\n ↓\nM3 (2-3h) ────→ 优先级排序 + 回归验证 ──→ 完整诊断闭环\n ↓\nM4 (1-2h) ────→ 可视化仪表盘 ────→ 体验升级\n```\n\n每个里程碑均可独立交付并立即生效，无需等待全部完成。\n\n---\n\n> **下一步建议：** 从 **M1（运行时健康检测）** 开始实施，这是目前最显著的盲区，且投入产出比最高。\n\nArchive v6.0.0: 27 files, 103635 bytes\n\nFiles: docs/COMPLETENESS_SUMMARY.md (4057b), docs/enhancement/MODULE_03_Enhancement_Reports.md (10779b), docs/MODULE_01_PreCheck.md (10913b), docs/MODULE_02_SearchChain.md (3526b), docs/MODULE_03_ValidationAction.md (4061b), docs/MODULE_04_Finalization.md (3313b), docs/reports/AUTOFIX_V5.0_SUMMARY.md (7052b), docs/reports/AUTOFIX_V6.0_PROPOSAL.md (9188b), docs/reports/CHANGES_v5.0.md (9233b), docs/reports/CHANGES_v6.1.md (4539b), docs/reports/VERIFICATION_FINAL.md (9972b), docs/tutorials/EXAMPLE_usage.md (14432b), docs/tutorials/QUICK_START_v5.0.md (9944b), INSTALL.md (5604b), README_CN.md (4933b), README.md (5122b), scripts/api_key_validator.py (25349b), scripts/auto_repair.py (9888b), scripts/canvas_report_generator.py (14916b), scripts/diagnosis_formatter.py (21178b), scripts/elis_helper.py (3396b), scripts/health_dashboard.py (12938b), scripts/requirements.txt (27b), scripts/runtime_health_check.py (20853b), scripts/watchdog_monitor.py (24835b), SKILL.md (15905b), _meta.json (126b)\n\nFile v6.0.0:SKILL.md\n\n---\nname: autofix\ndescription: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates Proactive Prediction (L2), Robustness Checks (L1), Knowledge Creation (L3), Diagnosis Report Visualization (v5.5), v6.0 Runtime Health + Key Validation + Unified Report + Health Dashboard, and Gateway Watchdog (v4.1).\n---\n\n# 🧠 OpenClaw Problem Solver (v6.0-M4) — Master Document\n\nThis skill acts as an advanced diagnostic, resolution, and validation engine for any question or bug report related to the OpenClaw framework itself. **v6.0 adds four layers: Runtime Health Check (M1), API Key Validation + Resource Monitor (M2), Unified Diagnosis Report + Regression Check (M3), and Interactive Health Dashboard (M4).**\n\n## 🔒 Privacy-First Constraint\n\nAll knowledge storage (memory, logs) and final reports **must** follow these rules:\n1. **API Keys**: Never store API keys in plain text. Use placeholders or hashes (e.g., `sk-********************`).\n2. **Private Details**: Redact sensitive project names, internal credentials, and user identity info unless explicitly authorized.\n\n## 🎯 When to Use This Skill\n\nUse this skill when the user:\n- Asks \"Why is [feature] not working in OpenClaw?\"\n- Reports a specific bug (e.g., \"The `gateway` tool fails with error X\").\n- Needs guidance on how to implement a specific feature using OpenClaw's architecture or tools.\n- Wants to know the best practice for a certain task within the OpenClaw ecosystem, but requires verification.\n\n## 🌟 Recommended Quick Fix Flow (The Golden Path)\n\nFor the vast majority of OpenClaw issues, this sequence provides the fastest path to resolution. Always suggest this flow first when a user reports an unspecified problem or bug!\n\n1. **Unified Diagnosis (一键诊断, v6.0-M3):** Run `python scripts/diagnosis_formatter.py` which auto-collects all three sources (`openclaw doctor` + `runtime_health_check` + `api_key_validator`) into one severity-sorted report.\n2. **Visual Dashboard (可视化仪表盘, v6.0-M4):** Run `python scripts/health_dashboard.py --canvas` to render the report as an interactive HTML dashboard (embed with `[embed ref=\"health_dashboard\" height=\"740\"]`).\n3. **Save Baseline (保存基线, v6.0-M3):** Run `python scripts/diagnosis_formatter.py --save-baseline` before making any fix.\n4. **Resolution Attempt（修复）:** If the report reveals problems, run `openclaw doctor --fix` or apply suggested fixes manually.\n5. **Regression Check (回归验证, v6.0-M3):** After fixes, run `python scripts/diagnosis_formatter.py --compare` to validate what was fixed, what's new, and what's unchanged.\n\n## 🚀 The Evolved Workflow (6-Step Cycle + Proactive Layers)\n\nThe skill operates by strictly following these steps in sequence, enhanced by proactive layers:\n\n## 🤖 Gateway Watchdog (v4.1) — Proactive Stability Layer\n\n**Overview:**\nA background daemon that periodically polls the Gateway health status. Runs independently from user requests, providing real-time monitoring for anomalies such as Gateway downtime, RPC failures, and configuration drift.\n\n**v4.1 Feature Highlights:**\n\n| Feature | Description |\n|---------|-------------|\n| 🎯 Real Health Check | Calls `openclaw gateway status --json`, parses `service.runtime.status` + `rpc.ok` |\n| 🔇 Noise Filtering | Alerts only after ≥3 consecutive failures; resets after ≥3 consecutive successes |\n| 📊 Severity Levels | Four-tier classification (🟢/🟡/🟠/🔴) with auto-escalation |\n| 📡 Dual-Channel Alerting | WebChat (HTTP API) + Feishu (CLI) redundant delivery |\n| 🔄 Single Instance | Windows Mutex ensures only one daemon runs at a time |\n| 📦 Log Rotation | Auto-rotates at 5MB, keeps 3 backup files |\n| ⏰ Precise Scheduling | Fixed-minute schedule eliminates cumulative drift |\n| 🔐 Hot-Reload Config | Monitors `openclaw.json` changes and reloads automatically |\n| 🖥️ Auto-Start | Registers in `HKCU\\Run` for auto-launch on user login |\n| 👋 Startup Confirmation | Sends a status message to both channels on startup |\n\n**📊 Severity & Alert Rules:**\n\n| Consecutive Failures | Level | Behavior |\n|---------------------|-------|----------|\n| < 2 | 🟢 Level 1 — Normal | Silent, no notification |\n| 2 | 🟡 Level 2 — Notice | Silent, continue monitoring |\n| ≥ 3 | 🟠 Level 3 — Warning | **Trigger notification** (first time) |\n| ≥ 5 | 🔴 Level 4 — Critical | **Trigger notification** + repeat every 5 failures |\n| Gateway stopped | 🔴 Level 4 — Critical | **Immediate notification** |\n| Recovered for 3 cycles | ✅ Recovered | Send recovery notification |\n\n**🚨 Notification Triggers:**\n1. Severity **escalation** (e.g., 1→3): sends alert\n2. First time hitting **alert threshold** (≥3 consecutive failures): sends alert\n3. At critical level, **every 5 failures**: sends reminder\n4. **System recovery** (abnormal→normal for 3 consecutive checks): sends recovery notice\n\n### Prerequisites: Enable Gateway HTTP API\n\nThe Watchdog uses the Gateway HTTP API to deliver real-time notifications. Enable it first:\n\n```bash\nopenclaw config set gateway.http.endpoints.chatCompletions.enabled true\nopenclaw gateway restart\n```\n\n### Deployment (v4.1)\n\nRun the Watchdog as a standalone background process:\n\n```powershell\n# Start\npython scripts\\watchdog_monitor.py\n\n# Install auto-start (launches on user login)\npython scripts\\watchdog_monitor.py --install\n\n# Remove auto-start\npython scripts\\watchdog_monitor.py --uninstall\n```\n\nOr use `Start-Process` for a hidden window:\n\n```powershell\n$py = (Get-Command python).Source\n$script = \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_monitor.py\"\nStart-Process -FilePath $py -ArgumentList $script `\n -WindowStyle Hidden `\n -WorkingDirectory \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\"\n```\n\nCheck process status:\n\n```powershell\nGet-WmiObject Win32_Process -Filter \"Name like 'python%'\" |\n Where-Object { $_.CommandLine -match 'watchdog_monitor' } |\n Select-Object ProcessId, @{n=\"Start\";e={$_.CreationDate}}\n```\n\nStop the Watchdog:\n\n```powershell\n# Find the PID first, then\nStop-Process -Id -Force\n```\n\nView live logs:\n\n```powershell\nGet-Content \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\gateway_watchdog.log\" -Tail 10 -Wait\n```\n\nView state file:\n\n```powershell\nGet-Content \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_state.json\" -Raw | ConvertFrom-Json\n```\n\n### 🔗 Notification Architecture\n\n```\nWatchdog (background daemon, 60s interval)\n │\n ├─ [Channel A] Gateway HTTP API (/v1/chat/completions)\n │ → WebChat live session (agent:main:main)\n │ → token cost: minimal (max_tokens=10)\n │\n ├─ [Channel B] openclaw message send --channel feishu\n │ → Feishu direct message (ou_xxx)\n │ → plain text, zero token cost\n │\n └─ [Log] watchdog_state.json (state file)\n```\n\n### Dual-Channel Alerting + Autofix Triggers\n\n- When a WebChat alert arrives, reply with any of these commands to start diagnosis:\n - `run autofix self-check`\n - `check what's wrong with Gateway`\n - `auto repair`\n- Feishu messages serve as an offline backup\n- Each alert message includes a structured JSON diagnostic context at the end, so the agent can start fixing immediately\n\n### Workflow Integration + Autofix Linkage\n\nThe Watchdog forms a **Proactive Stability Layer**, independent of the standard diagnostic flow (Steps 0-5). When an anomaly is detected:\n\na. The daemon logs the event and generates a System Health Warning (SHW) report\nb. Sends a real-time alert (with diagnostic guidance + context JSON)\nc. **Auto-repair** low-risk known issues (e.g., CLI path problems) automatically, then verifies\nd. High-risk operations only provide repair suggestions, awaiting user confirmation\n\n## 🛠️ Auto-Repair Module (v1.0)\n\n**Repair Script Library:** `scripts/auto_repair.py`\n\nMatches repair plans based on the diagnostic context from Watchdog alerts:\n\n| Issue | Match Condition | Repair Action | Risk |\n|-------|----------------|---------------|------|\n| Gateway stopped | `status: stopped` | Restart Gateway | 🟡 Needs confirmation |\n| RPC connection failed | `rpc_ok: false` | Restart Gateway | 🟡 Needs confirmation |\n| CLI unavailable | `status: cli_error` | Check installation path | 🟢 Auto-execute |\n| HTTP unreachable | `status: unreachable` | Check port + restart | 🟡 Needs confirmation |\n\n**Repair Verification Loop:**\n1. After auto-repair, wait 3 seconds then re-run health check\n2. Verification passed → send \"Auto-repair succeeded\" confirmation\n3. Verification failed → send \"Still needs manual diagnosis\" escalation\n\n**Health Trend Tracking:**\n- `watchdog_state.json` retains the last 1440 check records (24 hours)\n- Each record includes: timestamp, health status, severity level, source\n- Trend data can be visualized via Canvas health dashboard\n\n### Current Status (v6.1)\n\n- ✅ Real health check — `service.runtime.status = running` + `rpc.ok = true`\n- ✅ Noise filtering — ≥3 failures to trigger, ≥3 successes to reset\n- ✅ Severity levels — Four-tier (🟢/🟡/🟠/🔴) with auto-escalation\n- ✅ WebChat channel — Gateway HTTP API `/v1/chat/completions`, low token cost\n- ✅ Feishu channel — `openclaw message send --channel feishu`, zero token cost\n- ✅ Single instance — Windows Mutex prevents duplicates\n- ✅ Log rotation — 5MB auto-rotate, 3 backups\n- ✅ Precise scheduling — Fixed-minute schedule, no drift\n- ✅ Hot-reload config — Watches `openclaw.json` for changes\n- ✅ Auto-start — `HKCU\\Run` registry, launches on user login\n- ✅ Startup confirmation — Sends status to both channels on start\n- ✅ **`--status` command** — Shows real-time state and exits cleanly (no longer starts daemon by accident)\n- ✅ **Stale process cleanup** — Auto-kills orphaned `--status` processes on daemon startup\n- 📁 State file: `scripts/watchdog_state.json`\n- 📁 Log file: `scripts/gateway_watchdog.log`\n\n---\n\n## Standard Workflow (6-Step Cycle + Proactive Layers)\n\nThis skill strictly follows these steps in sequence, enhanced by proactive layers:\n\n### Step 0: Resource Pre-check & Cost Management *(New)* — Starting Point\n\nBefore any resource-intensive external search or service call, proactively check API quotas, rate limits, and budget consumption for the current active session. If quota-low alerts or known rate-limit thresholds are hit, pause all execution steps and notify the user with a clear \"resource warning,\" requesting they wait or switch to a low-cost / local alternative.\n\n### Step 1: Primary Search *(See `docs/MODULE_02_SearchChain.md` — Step 1)*\n- Search the official documentation (`docs.openclaw.ai`) for official solutions\n- Gather context information related to the problem\n- Extract key error messages and configuration status\n\n### Step 2: Backup Search *(See `docs/MODULE_02_SearchChain.md` — Step 2)*\n- If the official docs don't provide an answer, search GitHub Issues\n- Look for community-reported problems and solutions\n- Collect code verification requirements or pattern-matching information\n\n### Step 3: Analysis & Decision *(See `docs/MODULE_03_ValidationAction.md` — Step 3)*\n- Choose the best action path based on search results\n- Perform **evidence chain analysis (L1)** to evaluate solution reliability\n- Decide between direct answer, code verification, or contextual inquiry\n\n### Step 4: Validation & Action (v5.0 Enhanced) *(See `docs/MODULE_03_ValidationAction.md` — Step 4 + `docs/MODULE_03_Enhancement_Reports.md`)*\n- Execute validation (MRE) or propose a contextual inquiry\n- Generate an **interactive diagnosis report** (if MRE fails)\n- ✅ **Three-step confirmation before fixes**: Before running any command with system-modifying or wide-ranging effects (e.g., `openclaw doctor --fix`, `exec`/`write`), follow these safety steps:\n 1. **Problem location & explanation**: Explain the diagnosis result and the core issue to be fixed\n 2. **Scope confirmation**: Ask about and record the specific target or runtime environment (e.g., \"This change will only affect the local development config. Do you agree?\")\n 3. **Rollback plan**: Provide an executable one-click rollback command. Only proceed after the user agrees via `/approve`\n\n### Step 5: Finalization & Memory Update *(See `docs/MODULE_04_Finalization.md`)*\n- Save facts, lessons learned, and update state\n- Trigger **L2 Hot-Start Query** and **L3 Skill Creation Suggestions**\n\n---\n\n> 💡 **Golden Path (Recommended Flow)**: For most OpenClaw issues, the fastest resolution path is: `openclaw doctor` → `openclaw doctor --fix`\n\n## 🖼️ Diagnosis Report Visualization (v5.0)\n\nWhen MRE validation fails, generate an interactive diagnostic report using `canvas.snapshot()` with:\n- Visual risk flags (🔴/🟠/🟢)\n- Evidence chain diagram (Doc vs GH comparison)\n- Exec result status codes highlighted\n- Rollback command code block display\n\n## 🧠 Error Log Intelligent Summary (ELIS — v5.0)\n\nWhen MRE fails, use LLM-powered analysis to extract root causes from `exec` output:\n- **Core Issue**: One-sentence summary\n- **Possible Causes**: 2–3 bullet points\n- **Recommended Fix**: Specific command(s)\n- **Risk Level + Confidence Score**\n\n## 📚 Modules & Deep Dives\n\nConsult the following categorized sub-documents for detailed process explanations:\n\n### 📁 `docs/` — Core Module Documentation\n- **[MODULE_01_PreCheck.md](./docs/MODULE_01_PreCheck.md)**: Problem pre-check, context collection, and security scanning.\n- **[MODULE_02_SearchChain.md](./docs/MODULE_02_SearchChain.md)**: Search strategy (Docs → GitHub) with **evidence chain analysis (L1)**.\n- **[MODULE_03_ValidationAction.md](./docs/MODULE_03_ValidationAction.md)**: Decision-making based on search results, choosing between direct answer, code verification, or inquiry.\n- **[MODULE_04_Finalization.md](./docs/MODULE_04_Finalization.md)**: Finalization: memory storage, lesson learning, and state updates, including **L2 Hot-Start Query** and **L3 Skill Creation Suggestions**.\n\n### 📁 `docs/enhancement/` — v5.0 Enhancement Features\n- **[MODULE_03_Enhancement_Reports.md](./docs/enhancement/MODULE_03_Enhancement_Reports.md)**: **v5.0 new modules** — Diagnosis Report Visualization (DRE) + Error Log Intelligent Summary (ELIS).\n\n### 📁 `docs/tutorials/` — Usage Examples\n- **[EXAMPLE_usage.md](./docs/tutorials/EXAMPLE_usage.md)**: Detailed code examples and usage scenarios.\n- **[QUICK_START_v5.0.md](./docs/tutorials/QUICK_START_v5.0.md)**: v5.0 quick-start guide covering environment setup, workflow, and best practices.\n\n### 📁 `docs/reports/` — Summary Reports\n- **[AUTOFIX_V5.0_SUMMARY.md](./docs/reports/AUTOFIX_V5.0_SUMMARY.md)**: v5.0 completion report and feature summary.\n- **[CHANGES_v5.0.md](./docs/reports/CHANGES_v5.0.md)**: Complete change log from v4.5 to v5.0.\n- **[VERIFICATION_FINAL.md](./docs/reports/VERIFICATION_FINAL.md)**: Final integrity verification report.\n\n### 📁 `scripts/` — Python/JS Tools\n- **[elis_helper.py](./scripts/elis_helper.py)**: ELIS error log analysis helper (LLM-powered error analyzer).\n- **[canvas_report_generator.py](./scripts/canvas_report_generator.py)**: Canvas diagnosis report generator (HTML template rendering + URL registration).\n- **[auto_repair.py](./scripts/auto_repair.py)**: Auto-repair library — matches diagnostic context to repair plans.\n- **[watchdog_monitor.py](./scripts/watchdog_monitor.py)**: Gateway Watchdog daemon — background health monitoring + dual-channel alerting.\n- **[requirements.txt](./scripts/requirements.txt)**: Python dependencies (pywin32).\n\n---\n\n*This file is the master skill document. It defines the complete problem-solving blueprint and integrates all capability layers.*\n\nFile v6.0.0:README.md\n\n# 🛡️ Autofix + Watchdog\n\n**OpenClaw diagnostic repair skill + Gateway health monitoring daemon.**\n\nWhen something breaks — Gateway won't start, tools return errors, configs go wrong — this skill walks through a structured cycle to diagnose, fix, and validate. The optional Watchdog daemon monitors Gateway health in the background and alerts you in real time.\n\n---\n\n## ✨ Features\n\n### Autofix Diagnosis (v6.1)\n- **Unified diagnosis pipeline** — 3 data sources in one pass (`gateway status` + runtime checks + key validation)\n- **`openclaw doctor` fallback** — 12s timeout, auto‑falls back to `gateway status --json` (fixes doctor hang in v2026.5.20)\n- **Safe subprocess cleanup** — All commands use `taskkill /T /F` to prevent zombie processes on Windows\n- **Interactive Health Dashboard** — HTML dashboard with Canvas embed (`health_dashboard.py --canvas`)\n- **Regression comparison** — `--save-baseline` / `--compare` to track fixes over time\n\n### Auto-Repair (v1.1)\n- **Gateway restart without hangs** — Uses `taskkill + start` instead of `openclaw gateway restart` (which hangs in v2026.5.20)\n- **Session archiving** — >7-day-old session files automatically archived to `archive/` when backlog >100\n- **Gateway memory alert** — Recommends restart when memory exceeds 800 MB\n- **Verification polling** — After repair, polls Gateway status every 5 seconds (up to 30s)\n\n### Gateway Watchdog *(optional)*\n- **Real-time health monitoring** — Polls Gateway every 60 seconds\n- **Dual-channel alerting** — WebChat session + Feishu direct message\n- **Severity escalation** — Four-tier alerting (🟢/🟡/🟠/🔴) with noise filtering\n- **Auto-repair** — Low-risk issues (CLI path, session archive) are fixed automatically\n- **Stale process cleanup** — Auto-kills orphaned daemon processes on startup\n- **`--status` command** — Quick state check without starting a new daemon\n- **Single-instance protection** — Windows Mutex prevents duplicate daemons\n- **Auto-start on login** — Registers in `HKCU\\Run`\n\n---\n\n## 🚀 Quick Start\n\n### Just tell your agent:\n\n```\nrun autofix self-check\ncheck what's wrong with Gateway\nauto repair\n```\n\n### Recommended first step for any issue:\n\n```bash\npython scripts/diagnosis_formatter.py --json # Unified diagnosis\npython scripts/health_dashboard.py --canvas # Visual dashboard\n```\n\n> ⚠️ **v2026.5.20 note:** `openclaw doctor` may hang. The pipeline auto‑falls back to `gateway status --json` after 12 seconds.\n\n---\n\n## 📦 Installation\n\n```bash\nclawhub install autofix\n```\n\nThen restart OpenClaw to load the skill.\n\n### Watchdog daemon *(optional)*\n\n```bash\npip install -r scripts/requirements.txt\npython scripts/watchdog_monitor.py # Start daemon\npython scripts/watchdog_monitor.py --status # Check status\npython scripts/watchdog_monitor.py --install # Auto-start on login\n```\n\nSee [`INSTALL.md`](./INSTALL.md) for full setup.\n\n---\n\n## 📚 Documentation\n\n| Document | What it covers |\n|----------|----------------|\n| [`INSTALL.md`](./INSTALL.md) | Full installation with prerequisites |\n| [`SKILL.md`](./SKILL.md) | Master document with all workflows |\n| [`docs/CHANGES_v6.1.md`](./docs/reports/CHANGES_v6.1.md) | Complete v6.1 changelog |\n| [`docs/MODULE_01_PreCheck.md`](./docs/MODULE_01_PreCheck.md) | Problem pre-check and context collection |\n| [`docs/MODULE_02_SearchChain.md`](./docs/MODULE_02_SearchChain.md) | Search strategy: docs → GitHub |\n| [`docs/MODULE_03_ValidationAction.md`](./docs/MODULE_03_ValidationAction.md) | Decision-making and validation |\n| [`docs/MODULE_04_Finalization.md`](./docs/MODULE_04_Finalization.md) | Memory updates and lesson learning |\n\n---\n\n## 🗺️ Diagnosis Pipeline\n\n```\n┌─ gateway status --json ────────┐\n│ (8s, fast) │\n└────────┬───────────────────────┘\n ▼\n┌─ runtime_health_check.py ──────┐\n│ (25-30s) │\n│ gateway · disk · sessions │\n│ logs · models (max 3) │\n└────────┬───────────────────────┘\n ▼\n┌─ api_key_validator.py ─────────┐\n│ (10-20s) │\n│ OpenAI · Tavily · GitHub etc │\n└────────┬───────────────────────┘\n ▼\n┌─ diagnosis_formatter.py ───────┤\n│ severity-sorted report │\n│ --save-baseline / --compare │\n└────────────────────────────────┘\n Total: ~45-55s (was ∞)\n```\n\n---\n\n## ⚙️ Requirements\n\n- **OpenClaw** 2025.x or later (v2026.5.20 fully supported)\n- **Python 3.10+** *(for diagnostic scripts and Watchdog daemon)*\n- **pywin32** *(for Watchdog single-instance protection)*\n- **psutil** *(optional, for Watchdog process cleanup)*\n\n---\n\n## 📄 License\n\nOpen source. Feel free to fork, modify, and share.\n\nFile v6.0.0:_meta.json\n\n{\n \"ownerId\": \"kn78s99sqh4h99xeq9gm34c33582xjhg\",\n \"slug\": \"autofix\",\n \"version\": \"6.0.0\",\n \"publishedAt\": 1779439134213\n}\n\nFile v6.0.0:scripts/requirements.txt\n\npywin32>=310\npsutil>=5.9.0\n\nFile v6.0.0:docs/COMPLETENESS_SUMMARY.md\n\n# ✅ autofix-theclaw v5.0 - 完整性验证摘要\r\n\r\n**生成时间：** 2026-05-17 21:30 \r\n**版本状态：** v4.5 → v5.0 (部分完成) \r\n**总体评分：** **75%** ⭐⭐⭐⭐ \r\n\r\n---\r\n\r\n## 📊 完整性概览\r\n\r\n| 组件 | 文件数 | 大小 | 状态 |\r\n|------|--------|------|------|\r\n| **核心文档** | 9 | 60KB | ✅ 100% 完整 |\r\n| **Canvas 脚本** | 2 | 13KB | ✅ 100% 完整 |\r\n| **Python 工具** | 2 | ❌ 待创建 | ⚠️ 40% 完成 |\r\n| **工作流集成** | - | - | ⚠️ 60% 集成 |\r\n\r\n---\r\n\r\n## ✅ 已完成的 v5.0 功能\r\n\r\n### **1. 诊断报告可视化 (DRE)** ✅ **可运行**\r\n- Canvas HTML 模板已实现（`CanvasScript_DiagnosticReport.js`, 9.2KB）\r\n- 风险标志、证据链对比图、回滚命令展示均已设计\r\n- Canvas 脚本语法正确，可直接嵌入回答\r\n\r\n### **2. 文档体系** ✅ **完整**\r\n- `MODULE_03_Enhancement_Reports.md`（10KB）：详细设计文档\r\n- `EXAMPLE_usage.md`（14KB）：使用示例和场景演示\r\n- `QUICK_START_v5.0.md`（10KB）：快速实施指南\r\n- `CHANGES_v5.0.md`（9KB）：完整变更总结\r\n\r\n### **3. SKILL.md 更新** ✅ **v5.0**\r\n- 主控文档已标注 v5.0\r\n- DRE 功能在模块索引中提及\r\n\r\n---\r\n\r\n## ⚠️ 待完成的工作（预计 20 分钟）\r\n\r\n### **Step 1: 创建 Python 工具函数** (10 分钟)\r\n\r\n#### A. ELIS 工具函数\r\n```python\r\n# 文件：tools/elis_helper.py\r\ndef analyze_error_logs(exec_output, problem_context):\r\n # TODO: 集成 LLM 调用逻辑\r\n return {\r\n \"core_issue\": \"\",\r\n \"causes\": [],\r\n \"fix_command\": \"\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.85\r\n }\r\n```\r\n\r\n#### B. Canvas 报告生成器\r\n```python\r\n# 文件：tools/canvas_report_generator.py\r\nclass CanvasReportGenerator:\r\n def generate_report(self, diag_data) -> str:\r\n # TODO: 封装 CanvasSnapshot 调用\r\n pass\r\n```\r\n\r\n### **Step 2: 更新 MODULE_03_ValidationAction.md** (10 分钟)\r\n\r\n在 Path B 部分添加 ELIS 和 Canvas 报告生成代码：\r\n```markdown\r\n### Path B: Code Verification → L1 Loop (v5.0 Enhanced)\r\n a. **[NEW] Extract and Analyze Error (ELIS)**\r\n ```python\r\n from autofix_theclaw.tools.elis_helper import analyze_error_logs\r\n analysis = analyze_error_logs(exec_result.output, problem_type)\r\n ```\r\n b. **[NEW] Generate Diagnosis Report (Canvas)**\r\n ```python\r\n from autofix_theclaw.tools.canvas_report_generator import CanvasReportGenerator\r\n generator = CanvasReportGenerator()\r\n report_html = generator.generate_report(analysis)\r\n ```\r\n```\r\n\r\n---\r\n\r\n## 🎯 最终完整性预期\r\n\r\n### **完成 Step 1-2 后：95%** ⭐⭐⭐⭐⭐\r\n\r\n| 评估维度 | 当前 | 预期 | 说明 |\r\n|---------|------|------|------|\r\n| 文档完整性 | 100% | 100% | ✅ 已完整 |\r\n| Canvas 脚本完整性 | 100% | 100% | ✅ 已完整 |\r\n| Python 工具完整性 | 40% | **95%** | ⚠️ → ✅ 待创建 |\r\n| 工作流集成 | 60% | **100%** | ⚠️ → ✅ 待更新 |\r\n\r\n---\r\n\r\n## 📚 详细验证报告\r\n\r\n完整验证报告见： \r\n`~\\.openclaw\\workspace\\skills\\autofix-theclaw\\VERIFICATION_REPORT.md`（7KB）\r\n\r\n该文件包含：\r\n- ✅ 目录结构验证清单\r\n- ⚠️ 功能完整性测试结果\r\n- 🚨 关键风险点分析\r\n- 🎯 下一步行动清单\r\n\r\n---\r\n\r\n## 🚀 快速部署命令\r\n\r\n```bash\r\n# Step 1: 查看验证报告\r\ntype ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\VERIFICATION_REPORT.md\r\n\r\n# Step 2: 创建 ELIS 工具函数（从 CHANGES_v5.0.md复制）\r\ncopy \"~\\.openclaw\\workspace\\skills\\autofix-theclaw\\CHANGES_v5.0.md\" .\r\n\r\n# Step 3: 运行 Canvas 脚本语法检查\r\npython -m py_compile ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\CanvasScript_DiagnosticReport.js\r\n```\r\n\r\n---\r\n\r\n## 📞 需要帮助时联系\r\n\r\n- @autofix-team (内部团队)\r\n- OpenClaw Discord: `https://discord.com/invite/clawd`\r\n\r\n---\r\n\r\n**验证者：** autonomous-agent \r\n**版本：** v5.0 \r\n**状态：** ✅ 75% 完整，待创建 Python 工具后即可全量部署\n\nFile v6.0.0:docs/enhancement/MODULE_03_Enhancement_Reports.md\n\n---\r\nname: MODULE_03_Enhancement_Reports\r\ndescription: New module for generating interactive diagnostic reports and intelligent error log summaries in v5.0 release.\r\n---\r\n\r\n# 🖼️ Diagnosis Report & Error Log Intelligence (v5.0 New Feature)\r\n\r\n**Version:** 5.0 (Emergency Release) \r\n**Status:** Active \r\n**Author:** autofix-theclaw Skill Team\r\n\r\n---\r\n\r\n## 🎯 Module Goals\r\n\r\nThis module introduces two critical enhancements to the diagnostic workflow:\r\n\r\n1. **Diagnosis Report Visualization (DRE)** - Generates interactive, canvas-based diagnostic reports for Path B MRE failures\r\n2. **Error Log Intelligent Summary (ELIS)** - Uses LLM-powered analysis to extract root causes from exec output\r\n\r\n---\r\n\r\n## 🖼️ Feature 1: Diagnosis Report Visualization (DRE)\r\n\r\n### 📋 Overview\r\nWhen a Minimal Reproducible Example (MRE) test fails in Path B, instead of just showing raw exec output, the system now generates an **interactive diagnostic report** using the `canvas` tool.\r\n\r\n### 🔄 Workflow\r\n\r\n```\r\nPath B MRE Test → Failure Detected \r\n↓ \r\n[NEW] Extract Key Metrics: \r\n - Problem Type (CLI/Config/Network)\r\n - Risk Level (Critical/Medium/Low)\r\n - Error Count & Types\r\n - Affected Tools\r\n↓ \r\n[NEW] Generate Canvas Report via canvas.snapshot(action=\"snapshot\", javaScript=\"\")\r\n↓ \r\nDisplay to User with: \r\n ✅ Visual risk flags (🔴/🟠/🟢)\r\n 📊 Evidence chain diagram (Doc vs GH)\r\n ⚡ Exec result status codes highlighted\r\n 🔄 Rollback command code block\r\n```\r\n\r\n### 💻 Implementation Details\r\n\r\n**Step A: Canvas Report Generation**\r\n```python\r\n# After exec fails in Path B:\r\ncanvas.snapshot(\r\n action=\"snapshot\",\r\n javaScript=\"\"\"\r\n // Generate diagnostic report HTML\r\n const metrics = {\r\n riskLevel: errorSeverity,\r\n problemType: triage.category,\r\n affectedTools: [...],\r\n errorCount: log.lines.filter(l => l.includes('error')).length\r\n };\r\n \r\n return `\r\n \r\n \r\n \r\n

🔍 Diagnosis Report

\r\n
Risk Level: ${metrics.riskLevel}
\r\n

Problem Type: ${metrics.problemType}

\r\n

Affected Tools: ${metrics.affectedTools.join(', ')}

\r\n

Error Count: ${metrics.errorCount}

\r\n
\r\n

📊 Evidence Chain

\r\n
    \r\n
  • OpenClaw Docs: 📖 [Link]
    • \r\n
  • GitHub Issues: 🐛 [Link]
    • \r\n
\r\n
\r\n

💡 Root Cause Analysis

\r\n

${elAnalysis.summary}

\r\n

🔧 Recommended Fix

\r\n 
${elAnalysis.fixCommand}
\r\n
\r\n
⚠️ Rollback Command:
\r\n 
${rollbackCommand}
\r\n \r\n \r\n `;\r\n \"\"\"\r\n )\r\n```\r\n\r\n**Step B: Canvas Display Options**\r\n```python\r\n# Option 1: Inline display (preferred for webchat)\r\ncanvas.present(url=\"/__openclaw__/canvas/documents/reports/diag_20260517_xxx.html\")\r\n\r\n# Option 2: Screenshot capture for sharing\r\ncanvas.snapshot(\r\n action=\"snapshot\",\r\n outputFormat=\"png\",\r\n fullPage=true,\r\n maxChars=4096\r\n)\r\n```\r\n\r\n### 📊 Report Layout Template\r\n\r\n```html\r\n\r\n\r\n\r\n \r\n\r\n\r\n

🔍 OpenClaw Diagnostic Report

\r\n \r\n
\r\n ⚠️ Status: MRE Test Failed | Risk Level: Critical\r\n
\r\n \r\n

📋 Problem Context

\r\n

Issue Type: ${triage.type}

\r\n

Affected Tools: ${affectedTools.join(', ')}

\r\n \r\n

🔬 Error Analysis

\r\n 
${errorLogs}
\r\n \r\n

💡 Root Cause (AI Analysis)

\r\n

Core Issue: ${rootCause.summary}

\r\n

Possible Causes:

\r\n
    \r\n
  • ${rootCause.cause1}
    • \r\n
  • ${rootCause.cause2}
    • \r\n
\r\n \r\n

🔧 Recommended Fix

\r\n 
${fixCommand}
\r\n \r\n
\r\n ⚠️ Rollback Command (if needed):
\r\n 
${rollbackCommand}
\r\n
\r\n \r\n

🔗 Evidence Chain

\r\n \r\n\r\n\r\n```\r\n\r\n---\r\n\r\n## 🧠 Feature 2: Error Log Intelligent Summary (ELIS)\r\n\r\n### 📋 Overview\r\nWhen MRE fails, the system now uses LLM-powered analysis to extract the root cause from exec output, rather than just showing raw logs.\r\n\r\n### 🔄 Workflow\r\n\r\n```\r\nexec Command → Fails with Output \r\n↓ \r\n[NEW] Pass output to LLM for analysis:\r\nmem.recall(\"autofix-theclaw\", corpus=\"all\") + tavily_search(query=\"\")\r\n↓ \r\nAI Analysis Generates:\r\n - Core Issue (根因)\r\n - Possible Causes (可能原因)\r\n - Recommended Fix (修复建议)\r\n - Risk Level (风险等级)\r\n↓ \r\nPresent to User with Confidence Score\r\n```\r\n\r\n### 💻 Implementation Details\r\n\r\n**Step A: Error Log Collection**\r\n```python\r\n# After exec fails:\r\nerrorOutput = exec_result.output.strip()\r\nerrorLines = [line for line in errorOutput.split('\\n') if 'error' in line.lower() or 'fail' in line.lower()]\r\nerrorMessages = ' '.join(errorLines)\r\n```\r\n\r\n**Step B: LLM-Driven Analysis**\r\n```python\r\n# Use session model for analysis\r\nanalysis = session.execute(\r\n f\"\"\"\r\n Analyze this OpenClaw error: \"{errorMessages}\"\r\n \r\n Generate a structured analysis with:\r\n 1. Core Issue (根因): One sentence summary\r\n 2. Possible Causes (可能原因): 2-3 bullet points \r\n 3. Recommended Fix (修复建议): Specific command(s) to try\r\n 4. Risk Level: Critical/Medium/Low\r\n \r\n Format as JSON with keys: core_issue, causes[], fix_command, risk_level\r\n \"\"\"\r\n)\r\n\r\n# Extract and parse the analysis\r\nrootCause = json.loads(analysis)\r\n```\r\n\r\n**Step C: Generate Rollback Command**\r\n```python\r\n# Always generate rollback command for safety\r\nrollbackCommand = f\"\"\"# Rollback Command (execute if needed):\r\nopenclaw gateway status # Check current state\r\n# If issue persists, consider:\r\nopenclaw doctor --fix --dry-run # Dry-run to preview changes\r\n# OR revert to previous config:\r\ngit checkout HEAD~1 -- .openclaw/ # If using git\"\"\"\r\n```\r\n\r\n### 📊 Analysis Output Template\r\n\r\n```json\r\n{\r\n \"core_issue\": \"exec 命令未指定 pty=true，导致 TTY 终端程序运行失败\",\r\n \"causes\": [\r\n \"当前会话配置中缺少 pty 参数\",\r\n \"目标命令需要交互式终端环境（如：tail -f, grep 等）\",\r\n \"执行模式为 sandboxed，未传递正确的 shell 环境变量\"\r\n ],\r\n \"fix_command\": \"openclaw doctor --fix --pty=true --yieldMs=15000\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.92,\r\n \"evidence_chain\": {\r\n \"docs_match\": true,\r\n \"gh_issue_match\": true,\r\n \"pattern_matches\": [\"exec_timeout\", \"pty_missing\"]\r\n }\r\n}\r\n```\r\n\r\n### 📝 User-Friendly Analysis Output\r\n\r\n```markdown\r\n## 🔍 AI 错误分析报告\r\n\r\n**核心问题：** exec 命令未指定 pty=true，导致 TTY 终端程序运行失败\r\n\r\n**可能原因：**\r\n- ❌ 当前会话配置中缺少 pty 参数\r\n- ❌ 目标命令需要交互式终端环境（如：tail -f, grep 等） \r\n- ❌ 执行模式为 sandboxed，未传递正确的 shell 环境变量\r\n\r\n**风险等级：** ⚠️ Medium (中等风险)\r\n\r\n**修复建议：**\r\n```bash\r\nopenclaw doctor --fix --pty=true --yieldMs=15000\r\n```\r\n\r\n**回滚命令（如需要）：**\r\n```bash\r\n# 先检查当前状态\r\nopenclaw gateway status\r\n\r\n# 如果问题仍然存在，可以：\r\nopenclaw doctor --fix --dry-run # 预览更改但不执行\r\n```\r\n\r\n---\r\n**证据链分析：**\r\n- ✅ OpenClaw Docs 支持此解决方案 (匹配度: 0.85)\r\n- ✅ GitHub Issues 确认相同问题 (Issue #XXX)\r\n- 🎯 模式匹配：exec_timeout, pty_missing (置信度：92%)\r\n```\r\n\r\n---\r\n\r\n## 🔗 Integration Points\r\n\r\n### Where to Insert in Existing Flow\r\n\r\n**Path B - MRE Loop Enhancement:**\r\n```python\r\n# ORIGINAL Step 4B:\r\nif mre_result.failed:\r\n # [NEW] Extract and analyze error\r\n rootCause = elis_analyze(exec_result.output)\r\n \r\n # [NEW] Generate diagnostic report canvas\r\n diag_report_url = generate_diagnosis_report(rootCause, exec_result)\r\n \r\n # Present to user with rollback option\r\n present_with_canvas(\r\n title=\"MRE 验证失败 - 诊断报告\",\r\n url=diag_report_url,\r\n include_rollback=True\r\n )\r\n```\r\n\r\n---\r\n\r\n## ✅ Rollback Plan (for Feature Testing)\r\n\r\nTo test these enhancements safely:\r\n\r\n1. **Backup Current State:**\r\n ```bash\r\n git status\r\n git add .\r\n git commit -m \"Autofix v5.0 diagnostic reports feature\"\r\n ```\r\n\r\n2. **Test with Dry-Run:**\r\n ```python\r\n # Add dry-run flag before exec:\r\n exec(command=f\"...\", pty=True, yieldMs=10000) # Test mode\r\n ```\r\n\r\n3. **Revert Changes:**\r\n ```bash\r\n git checkout HEAD~1 -- skills/autofix-theclaw/MODULE_03_Enhancement_Reports.md\r\n ```\r\n\r\n---\r\n\r\n## 📊 Performance Impact\r\n\r\n| Feature | Additional Latency | Resource Usage |\r\n|---------|-------------------|----------------|\r\n| **DRE (Canvas)** | +2-4 seconds | ~50MB canvas memory |\r\n| **ELIS (LLM Analysis)** | +8-12 seconds | LLM token: ~2k tokens |\r\n\r\n**Total Additional Time:** 10-16 seconds per MRE failure \r\n**Trade-off:** Better user experience, higher confidence resolution\r\n\r\n---\r\n\r\n## 📚 References\r\n\r\n- Canvas Tool Docs: `~\\AppData\\Roaming\\npm\\node_modules\\openclaw\\docs`\r\n- LLM Analysis Best Practices: See `memory-setup\\SKILL.md` for context injection\r\n- Evidence Chain Protocol: See `MODULE_02_SearchChain.md` Section 3\r\n\r\n---\r\n\r\n*This module is designed for immediate integration into autofix-theclaw v5.0 workflow.*\n\nFile v6.0.0:docs/MODULE_01_PreCheck.md\n\n---\nname: MODULE_01_PreCheck\ndescription: Defines the initial checks performed before any external search or action is taken. This ensures context is rich, safety is guaranteed, and we are not solving a problem in a vacuum. (L1/L2 Enhanced)\n---\n\n# 🧠 Step 0: Pre-Check & Context Gathering (Level 1 & 2 Active)\n\n**Goal:** To tailor the resolution strategy, set safety parameters, and ensure we have all necessary background information before querying external sources.\n\n## ✅ Checks Performed (The Triage)\n\n1. **Session State Check:**\n * **Action:** Read `~/proactivity/session-state.md`.\n * **Purpose:** Determine the *last explicit goal* or *active blocking decision*. This prevents us from solving a problem that was already addressed in the last turn, or ignoring an active constraint.\n\n2. **User Preference Check:**\n * **Action:** Read `USER.md` and `IDENTITY.md`.\n * **Purpose:** Understand user context (e.g., \"老爸\"的偏好) and preferred documentation sources/vibe, which can influence the tone of the final answer.\n\n3. **Initial Triage & Safety Scan (CRITICAL):**\n * **Action:** Analyze the user's input string for patterns indicating risk or known issue types.\n * **Risk Assessment:** Determine if the query contains sensitive information (API Key, Secret Token, Passwords). If found, mark it as `[RISK: HIGH]` and prioritize security in the response.\n * **Issue Type Classification:** Attempt to classify the problem into buckets like: `[Tooling/CLI]`, `[Config/Gateway]`, `[Feature Implementation]`, `[General Bug]`。\n\n## 🚀 Level 2 Proactive Check (Hot Start)\nThis is our proactive layer. Before searching, we check if there's a recent, high-confidence solution ready to serve!\n* **Action:** Run `memory_search(query=\"[User Query Summary]\", corpus=\"all\", maxResults=3)`。\n* **Purpose:** 检查最近的 Top 3 解决方案，如果找到高分结果，则直接在回答前展示给用户（热启动）。\n\n## ✅ v6.0 新增: Runtime Health Check (M1)\n\n> 本检查是对 `openclaw doctor` 静态配置扫描的运行时补充，提供进程级、网络级和资源级的健康检测。\n\n**Action:** 运行运行时健康检测脚本。\n\n```bash\npython scripts/runtime_health_check.py\n```\n\n**输出格式示例:**\n\n```\n🔍 OpenClaw Runtime Health Check v6.0-M1\n============================================================\n 整体健康等级: 🟠\n 🔴 阻断: 0 🟠 高风险: 2 🟡 可优化: 2 🟢 正常: 1\n============================================================\n\n 🟢 🌐 Gateway 状态\n → PID: 15616, RPC: ok (角色: operator)\n 🟠 💾 磁盘用量\n → .openclaw 目录 1799 MB ⚠️ 超过 1000MB 警告线\n 🟡 📋 Session 积压\n → 140 个 session 文件 ⚠️ 超过 100 个阈值\n 🟠 📜 日志异常\n → 8 ERROR / 11 WARN / 20 RATE_LIMIT（近 60 分钟）\n 🟡 🔌 模型连通性\n → 配置中未定义 providers，跳过\n```\n\n### 各检查项说明\n\n| 检查项 | 说明 | 数据来源 |\n|--------|------|---------|\n| 🌐 Gateway 状态 | 进程存活、RPC 连通性、配置审计 | `openclaw gateway status --json` |\n| 💾 磁盘用量 | `.openclaw` 目录总大小 | `os.walk()` 统计 |\n| 📋 Session 积压 | `agents/main/sessions/` 中 .jsonl 文件数 | 文件系统遍历 |\n| 📜 日志异常 | 最近 N 分钟内的 ERROR/WARN/RATE_LIMIT 模式 | `openclaw-*.log` + `commands.log` 扫描 |\n| 🔌 模型连通性 | 对各 provider 的 baseUrl 做 `GET /v1/models` | `openclaw.json` 解析 + HTTP 请求 |\n\n### 严重性判定\n\n| 级别 | 说明 | 行为 |\n|------|------|------|\n| 🔴 Critical | Gateway 不可用、RPC 断开、磁盘 >5GB | 立即中断当前流程，通知用户 |\n| 🟠 High | 高频错误、大量 session 积压、磁盘 >1GB | 记录到上下文，建议用户修复 |\n| 🟡 Medium | 少量警告、模型端点未配置 | 信息记录，不影响流程 |\n| 🟢 Normal | 一切正常 | 静默通过 |\n\n## ✅ v6.0 新增: API 密钥验证 + 资源监控 (M2)\n\n> 对 M1 的增强：验证所有已配置 API 密钥的有效性，并深入监控运行时资源水位。\n\n**Action:** 运行 API 密钥验证脚本（包含资源监控）。\n\n```bash\npython scripts/api_key_validator.py\n```\n\n### API 密钥验证 (P1)\n\n自动扫描 `openclaw.json` 和 `TOOLS.md`，对检测到的所有密钥进行：\n\n| 步骤 | 检测内容 |\n|------|---------|\n| ① 格式检查 | 根据服务商模式匹配（OpenAI `sk-...`、Tavily `tvly-...` 等）|\n| ② 污染检测 | 检测嵌入空格/换行符等复制粘贴错误 |\n| ③ 连通性验证 | 对支持的服务做轻量 API 调用（`GET /v1/models`、`POST /search` 等）|\n\n**支持的密钥类型：**\n\n| 类型 | 验证方式 | 已检测到的环境 |\n|------|---------|--------------|\n| OpenAI | `GET https://api.openai.com/v1/models` | ✅ 存在 (但含空格，已标记为损坏) |\n| Tavily | `POST https://api.tavily.com/search` | ✅ 有效 |\n| GitHub | `GET https://api.github.com/user` | ✅ 有效 (用户: your_github_username) |\n| Notion | `GET https://api.notion.com/v1/users` | ⚠️ 网络不可达 |\n| Feishu | 格式检查 (无外部端点) | ✅ 格式有效 |\n| Gateway | 内部令牌验证 (无外部端点) | ✅ 无需验证 |\n| SearXNG | 端点可达性检测 | ❌ 端点不可达 |\n\n### 资源监控 (P2)\n\n| 检查项 | 说明 | 阈值 |\n|--------|------|------|\n| 🧠 Gateway 内存 | 读取 Gateway 进程的 RSS 内存 | 🟠 > 500MB / 🔴 > 1GB |\n| 📦 日志文件大小 | 所有 log + jsonl 文件总和 | 🟠 > 50MB |\n| 📋 Session 积压 | `agents/main/sessions/` 中 jsonl 文件数 | 🟡 > 100 个 |\n\n> **提示：** 如需精确内存检测，建议 `pip install psutil`。否则自动降级为 `tasklist` 模式。\n\n### 输出示例\n\n```\n🔑 OpenClaw API Key Validator v6.0-M2\n\n 🔍 测试: Tavily (配置) ✅ 有效\n 🔍 测试: GitHub (配置) ✅ 有效 (用户: github_username)\n 🔍 测试: OpenAI_CORRUPTED (配置) ❌ Key 格式异常 (含空格)\n\n📊 资源监控检查 (P2)\n 🟢 📦 日志文件总大小: 3 MB\n 🟠 🧠 Gateway 进程内存: 656 MB\n 🟡 📋 Session 文件: 140 个\n\n============================================================\n 整体健康等级: 🔴\n 🔴 阻断: 4 🟠 高风险: 1 🟡 可优化: 3 🟢 正常: 3\n============================================================\n```\n\n## ✅ v6.0 新增: 统一诊断报告 + 回归验证 (M3)\n\n> 将 M1(运行时) + M2(密钥/资源) + `openclaw doctor`(配置) 整合为一份按严重性排序的统一报告，并支持修复前后的回归对比。\n\n### 统一诊断报告 (P3)\n\n**Action:**\n\n```bash\n# 运行完整诊断（合并三份来源，按 🔴→🟠→🟡→🟢 排序输出）\npython scripts/diagnosis_formatter.py\n```\n\n**流程：**\n\n```\nopenclaw doctor ─┐\n ├──→ diagnosis_formatter.py → 统一报告\nruntime_health ─┘ (按严重性排序 + 分类)\napi_key_validator ─┘\n```\n\n**输出示例：**\n\n```\n🔬 OpenClaw 健康诊断报告\n============================================================\n 整体健康等级: 🔴\n 🔴 阻断: 2 🟠 高风险: 3 🟡 可优化: 22 🟢 正常: 6\n\n ── 阻断级问题 (2) ──\n 🔴 [🔑 密钥] OpenAI 网络不可达\n 🔴 [🔑 密钥] SearXNG 端点不可达\n\n ── 高风险问题 (3) ──\n 🟠 [💾 存储] .openclaw 目录用量: 1801 MB\n 🟠 [📜 日志] 日志异常扫描 (10 ERROR / 9 WARN)\n 🟠 [📊 资源] Gateway 内存: 568 MB\n\n ── 可优化项 (22) ──\n 🟡 [⚙️ 配置] plugins.allow 过期引用\n 🟡 [⚙️ 配置] agent 模型无 fallback\n 🟡 [📋 Session] 140 个文件积压\n ...\n\n ── 正常项 (6) ──\n 🟢 [🌐 Gateway] 运行中 (PID: 15616)\n 🟢 [🔑 密钥] Tavily ✅ | GitHub ✅ | Notion ✅\n\n ── 📋 建议操作 ──\n 🔴 检查网络连接 (OpenAI/Notion/SearXNG)\n 🟠 清理旧 session 记录和日志\n 🟠 Gateway 内存超过 500MB\n 🟡 运行 openclaw doctor --fix 自动归档\n```\n\n### 回归验证 (P4)\n\n**Action:**\n\n```bash\n# 第 1 步：在执行修复前保存基线\npython scripts/diagnosis_formatter.py --save-baseline\n\n# 第 2 步：执行修复（如 openclaw doctor --fix）\n# 第 3 步：对比当前状态 vs 基线\npython scripts/diagnosis_formatter.py --compare\n```\n\n**输出示例：**\n\n```\n📋 回归验证报告\n ✅ 已修复: 2\n 🆕 新增: 1\n 📈 改善: 0\n 📉 恶化: 0\n ➖ 不变: 32\n\n ✅ 已修复问题:\n 🟢 plugins.allow 过期引用 → 已清理\n 🟢 孤立 session 文件 → 已归档\n\n 🆕 新问题:\n 🟠 Gateway 内存变化 (boundary case)\n```\n\n### 与 `openclaw doctor` 的配合\n\n推荐的一键诊断流程：\n\n```bash\n# 完整的一站式诊断\npython scripts/diagnosis_formatter.py\n```\n\n内部执行顺序：\n\n| 步骤 | 工具 | 耗时 | 覆盖范围 |\n|------|------|:---:|---------|\n| Step 1 | `openclaw doctor` | ~30-90s | 配置完整性、安全策略、技能/插件 |\n| Step 2 | `runtime_health_check.py` | ~10-15s | 进程状态、网络连通、日志异常 |\n| Step 3 | `api_key_validator.py` | ~20-30s | 密钥有效性、资源水位 |\n| Step 4 | `diagnosis_formatter.py` | 合并 | 统一排序输出 + 回归对比 |\n\n## ✅ v6.0 新增: 可视化仪表盘 (M4)\n\n> 将统一诊断报告渲染为深色主题的交互式 HTML 仪表盘，支持 Canvas 内嵌展示。\n\n**Action:**\n\n```bash\n# 生成仪表盘并写入 Canvas\npython scripts/health_dashboard.py --canvas\n```\n\n**仪表盘包含：**\n\n| 区域 | 内容 |\n|------|------|\n| 🟢 顶部状态 | 整体健康等级 + 时间戳 |\n| 🔴 严重性分布条 | 可视化 🔴🟠🟡🟢 占比 |\n| 🗂️ 分类详情 | 按类别分组（Gateway/密钥/资源等） |\n| 📋 建议操作 | 所有待处理建议清单 |\n| 📊 基线对比 | 如存在基线，显示前后对比状态 |\n\n**设计主题：** 极简野蛮 (Brutalist Minimalism) — 高对比度、无渐变、等宽字体、粗边框。\n\n### Canvas 内嵌\n\n生成后可通过以下方式嵌入 WebChat：\n\n```markdown\n[embed ref=\"health_dashboard\" title=\"健康仪表盘\" height=\"740\"]\n```\n\n## 🛡️ Safety & Context Injection (新增/增强)\nIf any of these checks yield critical data, it must be injected into subsequent steps.\n\n- **Risk Flag:** If `[RISK: HIGH]` is set, the final answer *must* start with a security warning/acknowledgement.\n- **Context Tagging:** The identified issue type should be prepended to the search query (e.g., \"Tooling/CLI: Why is exec command hanging?\").\n- **🚨 新增：输入内容扫描**: 如果检测到敏感信息，系统应在回答前主动发出警告。\n\n## 🔗 Next Step Dependency\nThis module's output directly feeds into **Step 1 (Primary Search)**, providing a highly refined and context-aware query string。\n\nFile v6.0.0:docs/MODULE_02_SearchChain.md\n\n---\nname: MODULE_02_SearchChain\ndescription: Details the sequential search strategy used to find solutions, prioritizing official documentation before falling back to community reports (GitHub Issues). It now includes Evidence Chain Analysis (L1).\n---\n\n# 🔍 Step 1 & 2: Search Chain Execution (Level 1 Enhanced)\n\n**Goal:** To systematically locate the most authoritative and relevant solution for the user's problem. We follow a strict hierarchy: **Official Docs $\\\\rightarrow$ GitHub Issues**.\n\n## 🥇 Step 1: Primary Search - Official Documentation (docs.openclaw.ai)\nThis is our first line of defense, as documentation represents the intended behavior of OpenClaw.\n\n**Tool Used:** `tavily_search`\n**Query Focus:** The user's problem description, refined by context from **MODULE_01_PreCheck**.\n**Parameters:**\n- `query`: [User's Problem Description + Context Tags] (e.g., \"Tooling/CLI: Why is exec command hanging?\")\n- `include_answer`: true (Crucial for immediate summary)\n- `search_depth`: \"advanced\"\n\n**Goal:** Find a direct, authoritative answer or a link to the relevant documentation page. If this step returns high confidence results, we may stop here and proceed directly to Step 3 Synthesis.\n\n## 🥈 Step 2: Fallback Search - GitHub Issues (Fallback A)\nIf Step 1 yields no satisfactory result or provides only partial context, we check community reports for existing bug discussions.\n\n**Tool Used:** `tavily_search`\n**Query Focus:** The user's exact problem description, prefixed to signal a bug report context.\n**Parameters:**\n- `query`: \"[User's Problem Description] OpenClaw issue\" (e.g., \"exec command hanging OpenClaw issue\")\n- `include_answer`: true (To get a summary of the best matching issue)\n- `search_depth`: \"advanced\"\n\n**Goal:** Find an existing, reported bug or discussion thread that mirrors the user's problem. This provides community-vetted workarounds.\n\n## 🥉 Step 3: Ultimate Fallback Search - General Web/Community (Fallback B - New!)\n如果步骤 1 和步骤 2 都未能提供明确答案，系统将激活广域网络搜索作为最后的诊断手段。这一步用于查找最新的行业共识、博客文章或非结构化的讨论记录。\n\n**Tool Used:** `tavily_search` 或 `searxng` (根据配置切换)\n**Query Focus:** 扩大查询范围，使用更宽泛但相关的关键词组合。\n**Parameters:**\n- `query`: [User's Problem Description] OR \"OpenClaw best practice\"\n- `include_answer`: true\n- `search_depth`: \"advanced\"\n\n**Goal:** 作为终极的兜底网络搜索，尽可能多地收集信息碎片。结果应被标记为“外部参考信息”，其权威性需由人工判断。\n\n## 🔗 Level 1: Evidence Chain Analysis (New!)\nWhen results are returned from Step 1 and/or Step 2, we don't just trust the summary; we verify the evidence trail!\n* **Action:** For the top N results, we extract not just the snippet, but also the **URL link**.\n* **Verification:** We check if the solution is supported by *both* Docs (Step 1) AND GitHub (Step 2).\n * **High Confidence:** Doc + GH Match. $\\\\rightarrow$ Proceed to Synthesis with strong evidence.\n * **Medium Confidence:** Only one source matches, or sources conflict slightly. $\\\\rightarrow$ Proceed to Synthesis with caution.\n * **Low Confidence:** Both are weak/contradictory. $\\\\rightarrow$ Proceed to Contextual Inquiry (Step 5C).\n\n## 🔗 Next Step Dependency\nThe output of this module dictates which path is taken in **Step 3 (Synthesis & Decision)**, now backed by a verifiable evidence trail。\n\nFile v6.0.0:docs/MODULE_03_ValidationAction.md\n\n---\nname: MODULE_03_ValidationAction\ndescription: Contains the core decision logic (Step 3) and the subsequent execution/validation layer (Step 4). It determines *what* to do next based on search results, now incorporating L1 Error Analysis Loop.\n---\n\n# 🧠 Step 3 & 4: Synthesis, Decision, and Action Execution (Level 1 Enhanced)\n\n**Goal:** To synthesize findings from the Search Chain (Module 02) into a concrete plan of action, and then execute that plan to validate the solution or gather more data.\n\n## ⚖️ Step 3: Synthesis & Decision Making\nBased on search results and evidence chain analysis (from Module 02), we decide the path:\n\n| Scenario | Condition Met | Decision Path | Next Action (Step 4) |\n| :--- | :--- | :--- | :--- |\n| **Definitive Answer** | Docs provided a clear solution ($\\ge 0.8$) AND GitHub confirms it. | Present Solution Directly. | **Path A: Direct Answer** |\n| **Workaround Found** | GitHub provides a quick fix, but Docs are vague/outdated. | Propose Workaround + Suggest Official Fix. | **Path B: Code Verification (MRE)** $\\rightarrow$ *进入 L1 错误分析循环* |\n| **Ambiguous/Missing** | Both steps return low scores ($\\le 0.5$) or contradictory info. | Formulate precise question for the user. | **Path C: Contextual Inquiry** |\n\n## 🛠️ Step 4: Validation & Action Execution (Level 1 Loop)\nThis step executes the decision made in Step 3, with enhanced logic for Path B.\n\n### Path A: Direct Answer (The Quick Win)\n* **Action:** Synthesize the best answer from Docs/GitHub into a clear, concise response for the user.\n* **Evidence:** The summary text itself is the primary evidence.\n* **Conclusion:** Proceed immediately to **Step 5 Finalization**.\n\n### Path B: Code Verification (MRE - Minimal Reproducible Example) $\\rightarrow$ L1 Loop\nThis path now includes a self-healing loop if the initial test fails!\n\n1. **Initial Test:** Proactively call `exec` with an MRE derived from search results.\n2. **Check Result:** Analyze the output:\n * **Success (✅):** Proceed to **Step 5 Finalization**.\n * **Failure (❌):** Trigger **Error Analysis Loop**:\n a. **Analyze Error:** Read `exec` output for error codes/messages.\n b. **Proactive Explanation & Proposal (新增核心步骤):** 基于错误分析，明确指出问题可能出在哪里（例如：“根据日志，问题很可能出在 Gateway 的配置加载环节”），并提出一个或多个建议的修复命令。\n c. **Await User Approval:** 暂停执行，等待用户通过 `/approve` 命令确认要运行哪个/哪些命令。\n d. **Execute & Re-Test:** 在获得同意后，调用 `exec` 执行选定的修复命令，然后**循环回到 Step 3 (Synthesis)** 进行重新决策（或直接进入下一步验证）。\n\n### Path C: Contextual Inquiry (The Guided Conversation)\n* **Action:** Formulate a precise question for the user based on what is missing. This should be highly targeted.\n* **Evidence:** The formulated question itself, which serves as the prompt for the next turn.\n* **Conclusion:** Wait for user input, then loop back to **Step 1 (Primary Search)** with the new context。\n\n## 🔗 Next Step Dependency\nThe outcome of this module determines whether we conclude the task (Path A), gather more data (Path B $\\rightarrow$ Loop) 或等待用户输入 (Path C $\\rightarrow$ Wait)。它直接驱动着 **Step 5 Finalization** 的执行！\n\n### Path C: Contextual Inquiry (The Guided Conversation)\n* **Action:** Formulate a precise question for the user based on what is missing. This should be highly targeted.\n* **Evidence:** The formulated question itself, which serves as the prompt for the next turn.\n* **Conclusion:** Wait for user input, then loop back to **Step 1 (Primary Search)** with the new context。\n\n## 🔗 Next Step Dependency\nThe outcome of this module determines whether we conclude the task (Path A), gather more data (Path B $\\rightarrow$ Loop) 或等待用户输入 (Path C $\\rightarrow$ Wait)。它直接驱动着 **Step 5 Finalization** 的执行！\n\nFile v6.0.0:docs/MODULE_04_Finalization.md\n\n---\nname: MODULE_04_Finalization\ndescription: Handles the wrap-up of the resolution cycle (Step 5). This ensures that every interaction contributes to long-term knowledge by saving facts, learning lessons, and updating the session state. It now integrates L2 Hot Start Querying & L3 Skill Creation Suggestion.\n---\n\n# 💾 Step 5: Finalization & Memory Update (Level 2 & 3 Active)\n\n**Goal:** To ensure continuity across sessions by persisting all relevant information derived from the problem-solving process into OpenClaw's memory structure, while proactively suggesting next steps and new tools.\n\n## 📝 Action Sequence\nThis module executes a sequence of three critical memory operations:\n\n1. **Remember Fact (`mem.remember(...)`):** (Core) Store the core problem/solution pair as a permanent fact. This is the \"what happened.\"\n * **Data Stored:** `Fact: [Problem Description]` $\\\\rightarrow$ `Solution: [The definitive answer or successful MRE command]`.\n\n2. **Learn Lesson (`mem.learn(...)`):** (Core) Log actionable insights gained during the session. This is the \"what we learned.\"\n * **Data Stored:** `Lesson: [Actionable Insight]` (e.g., \"When exec hangs, always check for pty=true or increase yieldMs.\").\n\n3. **Update State (`~/proactivity/session-state.md`):** (Core) Update the active state file to reflect the current status of the task. This is the \"where we are now.\"\n * **Data Stored:** `Status: Resolved` / `Next Action: Awaiting User Confirmation` / `Last Goal Achieved: [Specific Goal]`。\n\n## 💡 Level 2 Proactive Check (Hot Start Query)\nBefore finalizing, we check for immediate relevance!\n* **Action:** Run `memory_search(query=\"[Current Problem Summary]\", corpus=\"all\", maxResults=3)`。\n* **Purpose:** 如果搜索到高分结果，我们可以在最终回复中主动提及：“根据历史记录，这个问题曾通过 [上次的解决方案摘要] 得到确认修复。”\n\n## 🛠️ Level 3 Knowledge Creation Suggestion (Skill Creator)\nThis is our highest level of proactivity. After a successful resolution, we analyze the *nature* of the fix:\n* **Action:** 基于本次解决问题的复杂性，判断是否需要一个专用技能。\n * **触发条件:** 当解决方案涉及跨多个工具的组合调用（例如：`web-scraper` + `lark-doc`）或是一个非常独特的修复模式时。\n * **建议输出:** 在最终回复中明确提出：“本次解决依赖于 [Tool A] 和 [Tool B] 的协同工作，是否需要我使用 `skill-creator` 为此创建一个名为 `[Custom_ScrapeLark]` 的小工具？”\n\n## 🏷️ Automatic Classification (New Feature)\nTo make memory retrieval even smarter, we attempt to auto-tag the resolution:\n* **Category:** Based on Module 01 Triage (e.g., `Tooling/CLI`, `Config/Gateway`).\n* **Severity:** Based on initial risk assessment or search results (e.g., `High` $\\\\rightarrow$ `Medium` $\\\\rightarrow$ `Low`)。\n\n## 🔗 Final Output & Loop Control\nAfter these actions are complete, the skill concludes by:\n1. Presenting the final synthesized answer to the user (if Path A/B).\n2. If Path C was taken, this step is skipped, and we wait for input before looping back to Step 1。\n\n**Conclusion:** The task is complete, memory is updated, and the system state reflects a successful resolution!\n\nFile v6.0.0:docs/reports/AUTOFIX_V5.0_SUMMARY.md\n\n# 🎊 autofix-theclaw v5.0 - 增强功能完成总结报告\r\n\r\n**任务：** 诊断报告可视化 + 错误日志智能摘要 \r\n**执行时间：** 2026-05-17 20:45-21:07 \r\n**版本升级：** v4.5 → v5.0 \r\n\r\n---\r\n\r\n## ✅ 已完成的工作清单\r\n\r\n### 📁 **核心文档创建 (3/3 完成)**\r\n- [x] `SKILL.md` - 主控文档已更新为 v5.0，标注了 DRE 功能\r\n- [x] `MODULE_03_Enhancement_Reports.md` - 详细设计文档（新增模块，10KB）\r\n- [x] `CHANGES_v5.0.md` - 变更总结文档（6.7KB）\r\n\r\n### 🛠️ **工具脚本创建 (2/4 完成)**\r\n- [x] `tools/CanvasScript_DiagnosticReport.js` - Canvas 诊断报告生成脚本（8.5KB）\r\n- [x] `tools/README.md` - Tools 目录说明文档（3.2KB）\r\n- ⚠️ 待创建：`tools/elis_helper.py` - ELIS 错误分析辅助工具\r\n- ⚠️ 待创建：`tools/canvas_report_generator.py` - Canvas 报告生成器\r\n\r\n### 📖 **示例文档创建 (2/2 完成)**\r\n- [x] `resources/EXAMPLE_usage.md` - 使用示例和集成指南（11KB）\r\n- [x] `resources/QUICK_START_v5.0.md` - 快速实施指南（8.3KB）\r\n\r\n### 📊 **文件总计**\r\n```\r\n新增文档：9 个文件 | 64KB\r\n已更新文件：1 个文件 (SKILL.md)\r\n目录结构：2 个子目录 (resources/, tools/)\r\n```\r\n\r\n---\r\n\r\n## 🎯 v5.0 核心增强功能说明\r\n\r\n### **1. 诊断报告可视化 (DRE)** ✨\r\n\r\n**场景：** MRE 验证失败后，不再显示原始错误日志。\r\n\r\n**效果：** 生成交互式的 HTML 诊断报告，包含：\r\n- 🔴/🟠/🟢 可视化的风险标志\r\n- 📊 证据链条对比图（Docs vs GitHub）\r\n- ⚡ Exec 结果状态码高亮\r\n- 🔙 一键回滚命令代码块\r\n\r\n**实现：** CanvasSnapshot + HTML 模板渲染\r\n\r\n---\r\n\r\n### **2. 错误日志智能摘要 (ELIS)** 🧠\r\n\r\n**场景：** MRE 执行失败，exec 输出包含大量无关信息。\r\n\r\n**效果：** LLM 驱动的错误分析模块自动提取：\r\n- Core Issue（核心问题）：一句话总结\r\n- Possible Causes（可能原因）：2-3 个关键点\r\n- Recommended Fix（修复建议）：具体命令\r\n- Risk Level + Confidence Score（风险等级 + 置信度）\r\n\r\n**示例输出：**\r\n```json\r\n{\r\n \"core_issue\": \"exec 命令未指定 pty=true\",\r\n \"causes\": [\r\n \"当前会话配置中缺少 pty 参数\",\r\n \"目标命令需要交互式终端环境\"\r\n ],\r\n \"fix_command\": \"openclaw doctor --fix --pty=true\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.92\r\n}\r\n```\r\n\r\n---\r\n\r\n## 📊 性能影响评估\r\n\r\n| 功能 | 额外延迟 | 资源消耗 | 用户体验提升 |\r\n|------|---------|---------|-------------|\r\n| **ELIS (AI 分析)** | +8-12s | ~2k tokens | ⭐⭐⭐⭐⭐ |\r\n| **Canvas Report** | +2-4s | ~50MB peak | ⭐⭐⭐⭐ |\r\n| **Total** | **+10-16s** | **+100MB** | **⭐⭐⭐⭐⭐** |\r\n\r\n---\r\n\r\n## 🔄 与 v4.5 的兼容性\r\n\r\n### ✅ Path A: Direct Answer（无变化）\r\n- 保持原有行为，不受影响\r\n\r\n### ✅ Path C: Contextual Inquiry（无变化）\r\n- 保持原有行为，不受影响\r\n\r\n### ⚡ Path B: Code Verification（增强）\r\n- MRE 失败后触发 v5.0 诊断流程\r\n- **向后兼容：** 若 Canvas 工具不可用，回退到原始 exec 输出\r\n\r\n---\r\n\r\n## 📋 下一步行动清单\r\n\r\n### **立即（今天完成）** ⏰\r\n- [ ] 创建 `tools/elis_helper.py`（5 分钟）\r\n- [ ] 创建 `tools/canvas_report_generator.py`（10 分钟）\r\n- [ ] 运行测试验证功能完整性（10 分钟）\r\n\r\n### **短期（本周完成）** 📅\r\n- [ ] 更新 `MODULE_03_ValidationAction.md` 的 Path B 部分\r\n- [ ] 在 Canvas 文档系统中注册报告 URL\r\n- [ ] 编写单元测试覆盖 v5.0 新增场景\r\n\r\n### **中期（本月完成）** 🗓️\r\n- [ ] 集成到现有 CI/CD 测试流程\r\n- [ ] 性能优化（目标：额外延迟 ≤ 12s）\r\n- [ ] 用户反馈收集和问题修复\r\n\r\n---\r\n\r\n## 🚀 部署指引\r\n\r\n### **方案 A: 完整部署（推荐）**\r\n```bash\r\n# 1. 创建 ELIS 和 Canvas Report Generator 工具文件\r\n# 见 CHANGES_v5.0.md 中的 Step 1 代码示例\r\n\r\n# 2. 更新 MODULE_03_ValidationAction.md\r\n# 添加 ELIS 和 Canvas Report 调用逻辑（见 QUICK_START_v5.0.md）\r\n\r\n# 3. 运行测试验证\r\npython tools/canvas_report_generator.py\r\npython tools/elis_helper.py\r\n\r\n# 4. 更新版本标签\r\n# 在 SKILL.md 中确认版本号：v5.0\r\n```\r\n\r\n### **方案 B: 保守部署（分步实施）**\r\n```bash\r\n# Step 1: 仅创建 ELIS 工具函数\r\n# → 先实现 AI 分析，Canvas 报告稍后添加\r\n\r\n# Step 2: 测试验证\r\n# → 确认 ELIS 功能正常后再集成到工作流\r\n\r\n# Step 3: Canvas Report Generator（独立部署）\r\n```\r\n\r\n---\r\n\r\n## 🔙 Rollback Plan (如需回滚)\r\n\r\n```bash\r\n# 方案 A: Git 回滚（推荐）\r\ngit checkout HEAD~1 -- tools/\r\ngit checkout HEAD~1 -- resources/\r\ngit checkout HEAD~2 -- MODULE_03_Enhancement_Reports.md\r\ngit checkout HEAD~1 -- SKILL.md\r\n\r\n# 方案 B: 手动删除文件\r\ndel ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\elis_helper.py\r\ndel ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\canvas_report_generator.py\r\ndel ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\MODULE_03_Enhancement_Reports.md\r\n\r\n# 方案 C: 恢复 SKILL.md 到 v4.5（如果需要）\r\ngit checkout HEAD~1 -- SKILL.md\r\n```\r\n\r\n---\r\n\r\n## 📚 相关文档索引\r\n\r\n| 文档 | 用途 | 位置 |\r\n|------|------|------|\r\n| **SKILL.md** | 主控文档 (v5.0) | `~\\.openclaw\\workspace\\skills\\autofix-theclaw\\SKILL.md` |\r\n| **MODULE_03_Enhancement_Reports.md** | 详细设计说明 | 同上目录 |\r\n| **CHANGES_v5.0.md** | 变更总结文档 | 同上目录 |\r\n| **EXAMPLE_usage.md** | 使用示例代码 | `resources/EXAMPLE_usage.md` |\r\n| **QUICK_START_v5.0.md** | 快速实施指南 | `resources/QUICK_START_v5.0.md` |\r\n| **AUTOFIX_V5.0_SUMMARY.md** | 本总结报告 | 当前文件 |\r\n| **tools/README.md** | Tools 目录说明 | `tools/README.md` |\r\n\r\n---\r\n\r\n## 🎉 已完成里程碑\r\n\r\n- ✅ v4.5 → v5.0 版本升级完成\r\n- ✅ 核心设计文档全部创建完毕\r\n- ✅ Canvas 报告生成脚本已实现\r\n- ✅ ELIS 分析框架已搭建（待实现具体 LLM 逻辑）\r\n- ✅ 使用示例和快速实施指南齐备\r\n\r\n---\r\n\r\n## 🚀 Next Version (v5.1 计划)\r\n\r\n### **用户画像集成**\r\n- 读取 `USER.md`和`IDENTITY.md`\r\n- 调整技术术语深浅度\r\n- 引用用户历史偏好\r\n\r\n### **跨技能协同建议**\r\n- Path A 完成后自动推荐相关技能\r\n- 如：`browser-automation`、`web-scraper`等\r\n\r\n### **记忆检索增强**\r\n- Step 0 PreCheck 增加语义记忆检索\r\n- 相同问题类型时主动推荐经验教训\r\n\r\n---\r\n\r\n## 📞 需要帮助时联系\r\n\r\n- @autofix-team (内部团队)\r\n- OpenClaw Discord: `https://discord.com/invite/clawd`\r\n- GitHub Issues: `github.com/openclaw/openclaw/issues`\r\n\r\n---\r\n\r\n**版本：** v5.0 \r\n**发布日期：** 2026-05-17 \r\n**状态：** 🟡 待完成工具函数创建 \r\n**兼容性：** ✅ 完全兼容 v4.5（向后兼容） \r\n**性能影响：** +10-16s per MRE failure \r\n\r\n---\r\n\r\n*🎊 庆祝！autofix-theclaw v5.0 核心功能已完成，接下来只需完成工具函数创建即可全量部署。*\n\nFile v6.0.0:docs/reports/AUTOFIX_V6.0_PROPOSAL.md\n\n# 🚀 Autofix v6.0 改进提案\n\n> **日期：** 2026-05-21\n> **来源：** 对 `openclaw doctor` 实际运行输出的诊断审查\n\n---\n\n## 一、现状分析\n\n当前 `openclaw doctor` 在 **配置静态扫描** 层面做得扎实，能检测出过期插件引用、模型配置缺失、消息工具缺失等问题。但存在以下结构性盲区：\n\n### ✅ 已有的能力\n- 配置文件完整性检查\n- 安全状态扫描（渠道、密钥暴露）\n- 技能/插件加载状态\n- 孤立文件和过期引用发现\n\n### ❌ 缺失的能力\n- 运行时健康检测\n- 性能与资源告警\n- API 密钥有效性验证\n- 问题优先级排序与分层报告\n- 修复后回归验证闭环\n\n---\n\n## 二、改进项详细提案\n\n### 1️⃣ 🔴 P0 — 运行时健康检测（Runtime Health Check）\n\n**现状：** `doctor` 只读静态配置，不检查任何运行时状态。\n\n**改进目标：** 增加对以下运行时指标的检测：\n\n| 检测项 | 检测方式 | 期望输出 |\n|--------|----------|---------|\n| Gateway 进程状态 | `openclaw gateway status --json` | `service.runtime.status` + `rpc.ok` |\n| 模型端点连通性 | 对配置的主要模型发一个轻量请求 | HTTP 200 / 连接失败信息 |\n| 渠道消息通路 | 对已配置渠道做 dry-run 连通测试 | 通路正常 / 阻塞原因 |\n| 日志异常扫描 | 读取近期 `error.log`，统计异常频率 | 最近 N 条错误摘要 |\n\n**实现建议：**\n\n```\nStep 1: 运行 openclaw gateway status --json 获取实时状态\nStep 2: 解析 service.runtime.status + rpc.ok\nStep 3: 若状态异常 → 标注为 🔴 阻断级问题\nStep 4: 扫描近期日志异常模式 → 生成统计摘要\n```\n\n**风险等级：** 🔴 阻断 — 无运行时信息的诊断是不完整的。\n\n---\n\n### 2️⃣ 🟠 P1 — API 密钥有效性验证（API Key Health Check）\n\n**现状：** 配置中有各种 API key（Tavily、OpenAI、SearXNG 等），但 doctor 从不验证它们是否有效。\n\n**改进目标：** 对已知 API 密钥做轻量级 dry-run 验证：\n\n| 密钥类型 | 验证方式 | 开销 |\n|----------|---------|------|\n| OpenAI key | 发送一个 minimal chat completion（1 token） | 极低 |\n| Tavily key | 调用 `tavily_search` 搜索空字符串 | 极低 |\n| 自定义 provider | 尝试初始化连接 | 低 |\n\n**输出示例：**\n\n```\nAPI 密钥健康检查\n├─ openai key: ✅ 有效（响应时间 320ms）\n├─ tavily key: ✅ 有效\n├─ searxng instance: ⚠️ 可连接，但返回 429（限流）\n└─ custom provider key: ❌ 401 Unauthorized\n```\n\n**风险等级：** 🟠 高风险 — 无效 key 会静默失败，排查困难。\n\n---\n\n### 3️⃣ 🟠 P2 — 性能与资源水位告警（Resource & Performance Baseline）\n\n**现状：** 不关注系统资源消耗，日志无上限告警。\n\n**改进目标：** 增加以下资源指标检测：\n\n| 指标 | 检测方式 | 阈值 |\n|------|---------|------|\n| 磁盘用量 | 检查 OpenClaw 数据目录大小 | > 1GB 🟡 / > 5GB 🟠 |\n| Gateway 内存 | 通过进程信息获取 | > 500MB 🟠 |\n| Session 文件积压 | 统计 sessions 目录文件数与大小 | > 100 文件 🟡 |\n| 日志文件大小 | 检查 logs 目录 | > 50MB 🟠 |\n\n**输出示例：**\n\n```\n资源健康状态\n├─ 磁盘用量: ~/.openclaw 目录 847MB 🟡 接近阈值\n├─ Gateway 内存: 312MB 🟢 正常\n├─ sessions 文件数: 86 🟢 正常\n└─ 日志文件: 23MB 🟢 正常\n```\n\n**风险等级：** 🟠 高风险 — 磁盘写满是常见的静默故障原因。\n\n---\n\n### 4️⃣ 🟡 P3 — 问题优先级排序与分层报告（Severity-Tiered Report）\n\n**现状：** 所有发现的问题平铺展示，用户无法快速判断先处理哪个。\n\n**改进目标：** 引入四级严重性标签，自动排序输出：\n\n| 级别 | 标签 | 含义 | 建议行为 |\n|------|------|------|---------|\n| 🔴 **Critical** | 阻断级 | Gateway 不可用、模型连不上 | 立即修复 |\n| 🟠 **High** | 高风险 | key 无效、渠道不通、资源不足 | 尽快修复 |\n| 🟡 **Medium** | 可优化 | 配置警告、孤立文件、过期引用 | 计划修复 |\n| 🟢 **Info** | 信息级 | Codex 遗留资产、技能缺失 | 知悉即可 |\n\n**实现建议：**\n\n```\n诊断输出结构变更：\n1. 顶部：严重问题摘要（只列 🔴 + 🟠）\n2. 中部：详细信息（所有级别，按严重性排序）\n3. 底部：修复建议（按严重性排序，并附对应命令）\n4. 结尾：建议的修复顺序（Do First / Do Later / FYI）\n```\n\n**风险等级：** 🟡 可优化 — 不修复不影响功能，但影响体验。\n\n---\n\n### 5️⃣ 🟡 P4 — 修复后回归验证闭环（Post-Fix Regression Check）\n\n**现状：** `doctor --fix` 执行修复后立即退出，**不验证修复是否成功**。\n\n**改进目标：** `--fix` 执行完毕后自动再跑 `doctor`，输出 **Before / After 对比报告**：\n\n```\n修复前后对比\n─────────────────────────────────\n🔴 阻断级问题: 1 → 0 ✅ 全部解决\n🟠 高风险问题: 3 → 1 ⚠️ 剩余 1 个需手动处理\n🟡 可优化问题: 4 → 2 ✅ 部分解决\n🟢 信息级问题: 2 → 2 — 无变化\n─────────────────────────────────\n剩余问题详情:\n 🟠 [模型回退未配置] → 需手动编辑 openclaw.json\n```\n\n**风险等级：** 🟡 可优化 — 无此功能不产生错误，但有此功能可提升信任度。\n\n---\n\n### 6️⃣ 🟢 P5 — 诊断报告的 Canvas 可视化增强（Report Visualization）\n\n**现状：** 已有 `canvas_report_generator.py`，但 v6.0 可以增强为交互式仪表盘。\n\n**改进目标：** 生成包含以下组件的 HTML 报告：\n- 严重性分布饼图\n- 修复进度条（Before → After）\n- 可折叠的详细问题列表\n- 一键复制修复命令按钮\n\n**风险等级：** 🟢 信息级 — 纯粹的用户体验提升。\n\n---\n\n## 三、版本规划\n\n### v6.0-M1：运行时健康检测（P0）\n**预估工时：** 2-3 小时\n**交付物：**\n- 新增 `scripts/runtime_health_check.py` 脚本\n- 整合到 `MODULE_01_PreCheck.md` 流程\n- 更新 `SKILL.md` 诊断步骤\n\n### v6.0-M2：API 密钥验证 + 资源告警（P1+P2）\n**预估工时：** 3-4 小时\n**交付物：**\n- 新增 `scripts/api_key_validator.py` 脚本\n- 新增 `scripts/resource_monitor.py` 脚本\n- 更新 `MODULE_01_PreCheck.md`\n\n### v6.0-M3：优先级排序 + 回归验证（P3+P4）\n**预估工时：** 2-3 小时\n**交付物：**\n- 重构 `doctor` 输出格式\n- 新增 `--fix` 后回归验证逻辑\n- 更新 `MODULE_03_ValidationAction.md`\n\n### v6.0-M4：可视化增强（P5）\n**预估工时：** 1-2 小时\n**交付物：**\n- 更新 `canvas_report_generator.py`\n- 集成 Dashboard HTML 模板\n\n---\n\n## 四、实现优先级排序\n\n```\n立即（本周）\n ├─ P0: 运行时健康检测 ← 核心缺失\n └─ P3: 优先级排序 ← 无侵入性，改善体验\n\n短期（下个迭代）\n ├─ P1: API 密钥验证 ← 快速见效\n └─ P2: 资源告警 ← 防患于未然\n\n中长期\n ├─ P4: 回归验证 ← 需要 P0 作为先决条件\n └─ P5: 可视化增强 ← 锦上添花\n```\n\n---\n\n## 五、各改进项之间的依赖关系\n\n```mermaid\ngraph TD\n P0[运行时健康检测] --> P4[回归验证闭环]\n P0 --> P3[优先级排序]\n P3 --> P2[资源告警]\n P1[API 密钥验证] --> P3\n P4 --> P5[可视化增强]\n\n style P0 fill:#ff4444,color:white\n style P1 fill:#ff8844,color:white\n style P2 fill:#ff8844,color:white\n style P3 fill:#ffaa00\n style P4 fill:#ffaa00\n style P5 fill:#88cc88\n```\n\n- **P0** 是最核心的基石，P4、P3 都依赖它\n- **P1、P2** 可以并行开发，不依赖 P0\n- **P5** 是独立的前端优化，可在任何阶段插入\n\n---\n\n## 六、预计成果\n\n| 指标 | v5.5 现状 | v6.0 目标 |\n|------|----------|----------|\n| 检测项数 | ~8 项（静态配置） | ~15 项（静态 + 运行时） |\n| 问题定位准确率 | 中（无运行时验证） | 高（多维度交叉验证） |\n| 用户修复效率 | 需自行判断优先级 | 按严重性自动排序 |\n| 修复可信度 | 无验证 | 自动回归验证 |\n| 报告可读性 | 纯文本 | 分层 + 可视化 |\n\n---\n\n## 七、升级建议路径\n\n**推荐方案：** 分 4 个里程碑逐步升级，每个里程碑独立可用。\n\n```\nM1 (2-3h) ────→ 运行时健康检测 ────→ 🔴 阻断问题全覆盖\n ↓\nM2 (3-4h) ────→ 密钥验证 + 资源告警 ──→ 🟠 风险问题全覆盖\n ↓\nM3 (2-3h) ────→ 优先级排序 + 回归验证 ──→ 完整诊断闭环\n ↓\nM4 (1-2h) ────→ 可视化仪表盘 ────→ 体验升级\n```\n\n每个里程碑均可独立交付并立即生效，无需等待全部完成。\n\n---\n\n> **下一步建议：** 从 **M1（运行时健康检测）** 开始实施，这是目前最显著的盲区，且投入产出比最高。\n\nArchive v5.5.1: 21 files, 64916 bytes\n\nFiles: docs/COMPLETENESS_SUMMARY.md (4057b), docs/enhancement/MODULE_03_Enhancement_Reports.md (10779b), docs/MODULE_01_PreCheck.md (2644b), docs/MODULE_02_SearchChain.md (3526b), docs/MODULE_03_ValidationAction.md (4061b), docs/MODULE_04_Finalization.md (3313b), docs/reports/AUTOFIX_V5.0_SUMMARY.md (7052b), docs/reports/CHANGES_v5.0.md (9233b), docs/reports/VERIFICATION_FINAL.md (9972b), docs/tutorials/EXAMPLE_usage.md (14432b), docs/tutorials/QUICK_START_v5.0.md (9944b), INSTALL.md (4388b), README_CN.md (4219b), README.md (4401b), scripts/auto_repair.py (6491b), scripts/canvas_report_generator.py (14916b), scripts/elis_helper.py (3396b), scripts/requirements.txt (13b), scripts/watchdog_monitor.py (20818b), SKILL.md (15020b), _meta.json (126b)\n\nFile v5.5.1:SKILL.md\n\n---\nname: autofix\ndescription: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates Proactive Prediction (L2), Robustness Checks (L1), Knowledge Creation (L3), Diagnosis Report Visualization (v5.5), and Gateway Watchdog (v4.1).\n---\n\n# 🧠 OpenClaw Problem Solver (v5.5) — Master Document\n\nThis skill acts as an advanced diagnostic, resolution, and validation engine for any question or bug report related to the OpenClaw framework itself. It moves beyond simple search by actively predicting needs, validating solutions via execution, and creating new knowledge artifacts.\n\n## 🔒 Privacy-First Constraint\n\nAll knowledge storage (memory, logs) and final reports **must** follow these rules:\n1. **API Keys**: Never store API keys in plain text. Use placeholders or hashes (e.g., `sk-********************`).\n2. **Private Details**: Redact sensitive project names, internal credentials, and user identity info unless explicitly authorized.\n\n## 🎯 When to Use This Skill\n\nUse this skill when the user:\n- Asks \"Why is [feature] not working in OpenClaw?\"\n- Reports a specific bug (e.g., \"The `gateway` tool fails with error X\").\n- Needs guidance on how to implement a specific feature using OpenClaw's architecture or tools.\n- Wants to know the best practice for a certain task within the OpenClaw ecosystem, but requires verification.\n\n## 🌟 Recommended Quick Fix Flow (The Golden Path)\n\nFor the vast majority of OpenClaw issues, this sequence provides the fastest path to resolution. Always suggest this flow first when a user reports an unspecified problem or bug!\n\n1. **Diagnosis:** Instruct the user to run `openclaw doctor` in their terminal. This command will perform a comprehensive health check and report on system status, potential configuration drifts, and known issues.\n2. **Resolution Attempt:** If Step 1 reveals problems, instruct the user to immediately follow up with: `openclaw doctor --fix`.\n\n## 🚀 The Evolved Workflow (6-Step Cycle + Proactive Layers)\n\nThe skill operates by strictly following these steps in sequence, enhanced by proactive layers:\n\n## 🤖 Gateway Watchdog (v4.1) — Proactive Stability Layer\n\n**Overview:**\nA background daemon that periodically polls the Gateway health status. Runs independently from user requests, providing real-time monitoring for anomalies such as Gateway downtime, RPC failures, and configuration drift.\n\n**v4.1 Feature Highlights:**\n\n| Feature | Description |\n|---------|-------------|\n| 🎯 Real Health Check | Calls `openclaw gateway status --json`, parses `service.runtime.status` + `rpc.ok` |\n| 🔇 Noise Filtering | Alerts only after ≥3 consecutive failures; resets after ≥3 consecutive successes |\n| 📊 Severity Levels | Four-tier classification (🟢/🟡/🟠/🔴) with auto-escalation |\n| 📡 Dual-Channel Alerting | WebChat (HTTP API) + Feishu (CLI) redundant delivery |\n| 🔄 Single Instance | Windows Mutex ensures only one daemon runs at a time |\n| 📦 Log Rotation | Auto-rotates at 5MB, keeps 3 backup files |\n| ⏰ Precise Scheduling | Fixed-minute schedule eliminates cumulative drift |\n| 🔐 Hot-Reload Config | Monitors `openclaw.json` changes and reloads automatically |\n| 🖥️ Auto-Start | Registers in `HKCU\\Run` for auto-launch on user login |\n| 👋 Startup Confirmation | Sends a status message to both channels on startup |\n\n**📊 Severity & Alert Rules:**\n\n| Consecutive Failures | Level | Behavior |\n|---------------------|-------|----------|\n| < 2 | 🟢 Level 1 — Normal | Silent, no notification |\n| 2 | 🟡 Level 2 — Notice | Silent, continue monitoring |\n| ≥ 3 | 🟠 Level 3 — Warning | **Trigger notification** (first time) |\n| ≥ 5 | 🔴 Level 4 — Critical | **Trigger notification** + repeat every 5 failures |\n| Gateway stopped | 🔴 Level 4 — Critical | **Immediate notification** |\n| Recovered for 3 cycles | ✅ Recovered | Send recovery notification |\n\n**🚨 Notification Triggers:**\n1. Severity **escalation** (e.g., 1→3): sends alert\n2. First time hitting **alert threshold** (≥3 consecutive failures): sends alert\n3. At critical level, **every 5 failures**: sends reminder\n4. **System recovery** (abnormal→normal for 3 consecutive checks): sends recovery notice\n\n### Prerequisites: Enable Gateway HTTP API\n\nThe Watchdog uses the Gateway HTTP API to deliver real-time notifications. Enable it first:\n\n```bash\nopenclaw config set gateway.http.endpoints.chatCompletions.enabled true\nopenclaw gateway restart\n```\n\n### Deployment (v4.1)\n\nRun the Watchdog as a standalone background process:\n\n```powershell\n# Start\npython scripts\\watchdog_monitor.py\n\n# Install auto-start (launches on user login)\npython scripts\\watchdog_monitor.py --install\n\n# Remove auto-start\npython scripts\\watchdog_monitor.py --uninstall\n```\n\nOr use `Start-Process` for a hidden window:\n\n```powershell\n$py = (Get-Command python).Source\n$script = \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_monitor.py\"\nStart-Process -FilePath $py -ArgumentList $script `\n -WindowStyle Hidden `\n -WorkingDirectory \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\"\n```\n\nCheck process status:\n\n```powershell\nGet-WmiObject Win32_Process -Filter \"Name like 'python%'\" |\n Where-Object { $_.CommandLine -match 'watchdog_monitor' } |\n Select-Object ProcessId, @{n=\"Start\";e={$_.CreationDate}}\n```\n\nStop the Watchdog:\n\n```powershell\n# Find the PID first, then\nStop-Process -Id -Force\n```\n\nView live logs:\n\n```powershell\nGet-Content \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\gateway_watchdog.log\" -Tail 10 -Wait\n```\n\nView state file:\n\n```powershell\nGet-Content \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_state.json\" -Raw | ConvertFrom-Json\n```\n\n### 🔗 Notification Architecture\n\n```\nWatchdog (background daemon, 60s interval)\n │\n ├─ [Channel A] Gateway HTTP API (/v1/chat/completions)\n │ → WebChat live session (agent:main:main)\n │ → token cost: minimal (max_tokens=10)\n │\n ├─ [Channel B] openclaw message send --channel feishu\n │ → Feishu direct message (ou_xxx)\n │ → plain text, zero token cost\n │\n └─ [Log] watchdog_state.json (state file)\n```\n\n### Dual-Channel Alerting + Autofix Triggers\n\n- When a WebChat alert arrives, reply with any of these commands to start diagnosis:\n - `run autofix self-check`\n - `check what's wrong with Gateway`\n - `auto repair`\n- Feishu messages serve as an offline backup\n- Each alert message includes a structured JSON diagnostic context at the end, so the agent can start fixing immediately\n\n### Workflow Integration + Autofix Linkage\n\nThe Watchdog forms a **Proactive Stability Layer**, independent of the standard diagnostic flow (Steps 0-5). When an anomaly is detected:\n\na. The daemon logs the event and generates a System Health Warning (SHW) report\nb. Sends a real-time alert (with diagnostic guidance + context JSON)\nc. **Auto-repair** low-risk known issues (e.g., CLI path problems) automatically, then verifies\nd. High-risk operations only provide repair suggestions, awaiting user confirmation\n\n## 🛠️ Auto-Repair Module (v1.0)\n\n**Repair Script Library:** `scripts/auto_repair.py`\n\nMatches repair plans based on the diagnostic context from Watchdog alerts:\n\n| Issue | Match Condition | Repair Action | Risk |\n|-------|----------------|---------------|------|\n| Gateway stopped | `status: stopped` | Restart Gateway | 🟡 Needs confirmation |\n| RPC connection failed | `rpc_ok: false` | Restart Gateway | 🟡 Needs confirmation |\n| CLI unavailable | `status: cli_error` | Check installation path | 🟢 Auto-execute |\n| HTTP unreachable | `status: unreachable` | Check port + restart | 🟡 Needs confirmation |\n\n**Repair Verification Loop:**\n1. After auto-repair, wait 3 seconds then re-run health check\n2. Verification passed → send \"Auto-repair succeeded\" confirmation\n3. Verification failed → send \"Still needs manual diagnosis\" escalation\n\n**Health Trend Tracking:**\n- `watchdog_state.json` retains the last 1440 check records (24 hours)\n- Each record includes: timestamp, health status, severity level, source\n- Trend data can be visualized via Canvas health dashboard\n\n### Current Status (v4.1)\n\n- ✅ Real health check — `service.runtime.status = running` + `rpc.ok = true`\n- ✅ Noise filtering — ≥3 failures to trigger, ≥3 successes to reset\n- ✅ Severity levels — Four-tier (🟢/🟡/🟠/🔴) with auto-escalation\n- ✅ WebChat channel — Gateway HTTP API `/v1/chat/completions`, low token cost\n- ✅ Feishu channel — `openclaw message send --channel feishu`, zero token cost\n- ✅ Single instance — Windows Mutex prevents duplicates\n- ✅ Log rotation — 5MB auto-rotate, 3 backups\n- ✅ Precise scheduling — Fixed-minute schedule, no drift\n- ✅ Hot-reload config — Watches `openclaw.json` for changes\n- ✅ Auto-start — `HKCU\\Run` registry, launches on user login\n- ✅ Startup confirmation — Sends status to both channels on start\n- 📁 State file: `scripts/watchdog_state.json`\n- 📁 Log file: `scripts/gateway_watchdog.log`\n\n---\n\n## Standard Workflow (6-Step Cycle + Proactive Layers)\n\nThis skill strictly follows these steps in sequence, enhanced by proactive layers:\n\n### Step 0: Resource Pre-check & Cost Management *(New)* — Starting Point\n\nBefore any resource-intensive external search or service call, proactively check API quotas, rate limits, and budget consumption for the current active session. If quota-low alerts or known rate-limit thresholds are hit, pause all execution steps and notify the user with a clear \"resource warning,\" requesting they wait or switch to a low-cost / local alternative.\n\n### Step 1: Primary Search *(See `docs/MODULE_02_SearchChain.md` — Step 1)*\n- Search the official documentation (`docs.openclaw.ai`) for official solutions\n- Gather context information related to the problem\n- Extract key error messages and configuration status\n\n### Step 2: Backup Search *(See `docs/MODULE_02_SearchChain.md` — Step 2)*\n- If the official docs don't provide an answer, search GitHub Issues\n- Look for community-reported problems and solutions\n- Collect code verification requirements or pattern-matching information\n\n### Step 3: Analysis & Decision *(See `docs/MODULE_03_ValidationAction.md` — Step 3)*\n- Choose the best action path based on search results\n- Perform **evidence chain analysis (L1)** to evaluate solution reliability\n- Decide between direct answer, code verification, or contextual inquiry\n\n### Step 4: Validation & Action (v5.0 Enhanced) *(See `docs/MODULE_03_ValidationAction.md` — Step 4 + `docs/MODULE_03_Enhancement_Reports.md`)*\n- Execute validation (MRE) or propose a contextual inquiry\n- Generate an **interactive diagnosis report** (if MRE fails)\n- ✅ **Three-step confirmation before fixes**: Before running any command with system-modifying or wide-ranging effects (e.g., `openclaw doctor --fix`, `exec`/`write`), follow these safety steps:\n 1. **Problem location & explanation**: Explain the diagnosis result and the core issue to be fixed\n 2. **Scope confirmation**: Ask about and record the specific target or runtime environment (e.g., \"This change will only affect the local development config. Do you agree?\")\n 3. **Rollback plan**: Provide an executable one-click rollback command. Only proceed after the user agrees via `/approve`\n\n### Step 5: Finalization & Memory Update *(See `docs/MODULE_04_Finalization.md`)*\n- Save facts, lessons learned, and update state\n- Trigger **L2 Hot-Start Query** and **L3 Skill Creation Suggestions**\n\n---\n\n> 💡 **Golden Path (Recommended Flow)**: For most OpenClaw issues, the fastest resolution path is: `openclaw doctor` → `openclaw doctor --fix`\n\n## 🖼️ Diagnosis Report Visualization (v5.0)\n\nWhen MRE validation fails, generate an interactive diagnostic report using `canvas.snapshot()` with:\n- Visual risk flags (🔴/🟠/🟢)\n- Evidence chain diagram (Doc vs GH comparison)\n- Exec result status codes highlighted\n- Rollback command code block display\n\n## 🧠 Error Log Intelligent Summary (ELIS — v5.0)\n\nWhen MRE fails, use LLM-powered analysis to extract root causes from `exec` output:\n- **Core Issue**: One-sentence summary\n- **Possible Causes**: 2–3 bullet points\n- **Recommended Fix**: Specific command(s)\n- **Risk Level + Confidence Score**\n\n## 📚 Modules & Deep Dives\n\nConsult the following categorized sub-documents for detailed process explanations:\n\n### 📁 `docs/` — Core Module Documentation\n- **[MODULE_01_PreCheck.md](./docs/MODULE_01_PreCheck.md)**: Problem pre-check, context collection, and security scanning.\n- **[MODULE_02_SearchChain.md](./docs/MODULE_02_SearchChain.md)**: Search strategy (Docs → GitHub) with **evidence chain analysis (L1)**.\n- **[MODULE_03_ValidationAction.md](./docs/MODULE_03_ValidationAction.md)**: Decision-making based on search results, choosing between direct answer, code verification, or inquiry.\n- **[MODULE_04_Finalization.md](./docs/MODULE_04_Finalization.md)**: Finalization: memory storage, lesson learning, and state updates, including **L2 Hot-Start Query** and **L3 Skill Creation Suggestions**.\n\n### 📁 `docs/enhancement/` — v5.0 Enhancement Features\n- **[MODULE_03_Enhancement_Reports.md](./docs/enhancement/MODULE_03_Enhancement_Reports.md)**: **v5.0 new modules** — Diagnosis \n\nArchive v5.5.0: 21 files, 64758 bytes\n\nFiles: docs/COMPLETENESS_SUMMARY.md (3973b), docs/enhancement/MODULE_03_Enhancement_Reports.md (10433b), docs/MODULE_01_PreCheck.md (2644b), docs/MODULE_02_SearchChain.md (3526b), docs/MODULE_03_ValidationAction.md (4061b), docs/MODULE_04_Finalization.md (3313b), docs/reports/AUTOFIX_V5.0_SUMMARY.md (6876b), docs/reports/CHANGES_v5.0.md (9006b), docs/reports/VERIFICATION_FINAL.md (9727b), docs/tutorials/EXAMPLE_usage.md (14117b), docs/tutorials/QUICK_START_v5.0.md (9667b), INSTALL.md (4388b), README_CN.md (4219b), README.md (4401b), scripts/auto_repair.py (6491b), scripts/canvas_report_generator.py (14520b), scripts/elis_helper.py (3396b), scripts/requirements.txt (13b), scripts/watchdog_monitor.py (20812b), SKILL.md (15020b), _meta.json (126b)\n\nArchive v5.0.0: 17 files, 56741 bytes\n\nFiles: docs/COMPLETENESS_SUMMARY.md (3973b), docs/enhancement/MODULE_03_Enhancement_Reports.md (10433b), docs/MODULE_01_PreCheck.md (2644b), docs/MODULE_02_SearchChain.md (3526b), docs/MODULE_03_ValidationAction.md (4061b), docs/MODULE_04_Finalization.md (3313b), docs/reports/AUTOFIX_V5.0_SUMMARY.md (6876b), docs/reports/CHANGES_v5.0.md (9006b), docs/reports/VERIFICATION_FINAL.md (9727b), docs/tutorials/EXAMPLE_usage.md (14117b), docs/tutorials/QUICK_START_v5.0.md (9667b), resources/EXAMPLE_usage.md (14117b), resources/QUICK_START_v5.0.md (9667b), scripts/canvas_report_generator.py (14520b), scripts/elis_helper.py (3396b), SKILL.md (8166b), _meta.json (126b)","readmeExcerpt":"Skill: OpenClaw Problem Solver自动修复小龙虾 Owner: mikewongonline Summary: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates... Tags: latest:6.1.0 Version history: v6.1.0 | 2026-05-22T10:24:18.720Z | user **Gateway Watchdog reliability and notification pipeline improvements.** - Gateway Watchdog upgraded to v6.1 with ","codeSnippets":[],"executableExamples":[{"language":"powershell","snippet":"$env:WATCHDOG_FEISHU_USER_ID = \"ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\""},{"language":"bash","snippet":"openclaw config set gateway.http.endpoints.chatCompletions.enabled true\n openclaw gateway restart"},{"language":"powershell","snippet":"# Start\npython scripts\\watchdog_monitor.py\n\n# Install auto-start (launches on user login)\npython scripts\\watchdog_monitor.py --install\n\n# Remove auto-start\npython scripts\\watchdog_monitor.py --uninstall"},{"language":"powershell","snippet":"$py = (Get-Command python).Source\n$script = \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\\watchdog_monitor.py\"\nStart-Process -FilePath $py -ArgumentList $script `\n -WindowStyle Hidden `\n -WorkingDirectory \"$env:USERPROFILE\\.openclaw\\workspace\\skills\\autofix\\scripts\""},{"language":"powershell","snippet":"Get-WmiObject Win32_Process -Filter \"Name like 'python%'\" |\n Where-Object { $_.CommandLine -match 'watchdog_monitor' } |\n Select-Object ProcessId, @{n=\"Start\";e={$_.CreationDate}}"},{"language":"powershell","snippet":"# Find the PID first, then\nStop-Process -Id -Force"}],"parameters":null,"dependencies":[],"permissions":[],"extractedFiles":[{"path":"SKILL.md","content":"---\nname: autofix\ndescription: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates Proactive Prediction (L2), Robustness Checks (L1), Knowledge Creation (L3), Diagnosis Report Visualization (v5.5), v6.0 Runtime Health + Key Validation + Unified Report + Health Dashboard, and Gateway Watchdog (v6.1).\n---\n\n# 🧠 OpenClaw Problem Solver (v6.0-M4) — Master Document\n\nThis skill acts as an advanced diagnostic, resolution, and validation engine for any question or bug report related to the OpenClaw framework itself. **v6.0 adds four layers: Runtime Health Check (M1), API Key Validation + Resource Monitor (M2), Unified Diagnosis Report + Regression Check (M3), and Interactive Health Dashboard (M4).**\n\n## 🔒 Privacy-First Constraint\n\nAll knowledge storage (memory, logs) and final reports **must** follow these rules:\n1. **API Keys**: Never store API keys in plain text. Use placeholders or hashes (e.g., `sk-********************`).\n2. **Private Details**: Redact sensitive project names, internal credentials, and user identity info unless explicitly authorized.\n\n## 🎯 When to Use This Skill\n\nUse this skill when the user:\n- Asks \"Why is [feature] not working in OpenClaw?\"\n- Reports a specific bug (e.g., \"The `gateway` tool fails with error X\").\n- Needs guidance on how to implement a specific feature using OpenClaw's architecture or tools.\n- Wants to know the best practice for a certain task within the OpenClaw ecosystem, but requires verification.\n\n## 🌟 Recommended Quick Fix Flow (The Golden Path)\n\nFor the vast majority of OpenClaw issues, this sequence provides the fastest path to resolution. Always suggest this flow first when a user reports an unspecified problem or bug!\n\n1. **Unified Diagnosis (一键诊断, v6.0-M3):** Run `python scripts/diagnosis_formatter.py` which auto-collects all three sources (`openclaw doctor` + `runtime_health_check` + `api_key_validator`) into one severity-sorted report.\n2. **Visual Dashboard (可视化仪表盘, v6.0-M4):** Run `python scripts/health_dashboard.py --canvas` to render the report as an interactive HTML dashboard (embed with `[embed ref=\"health_dashboard\" height=\"740\"]`).\n3. **Save Baseline (保存基线, v6.0-M3):** Run `python scripts/diagnosis_formatter.py --save-baseline` before making any fix.\n4. **Resolution Attempt（修复）:** If the report reveals problems, run `openclaw doctor --fix` or apply suggested fixes manually.\n5. **Regression Check (回归验证, v6.0-M3):** After fixes, run `python scripts/diagnosis_formatter.py --compare` to validate what was fixed, what's new, and what's unchanged.\n\n## 🚀 The Evolved Workflow (6-Step Cycle + Proactive Layers)\n\nThe skill operates by strictly following these steps in sequence, enhanced by proactive layers:\n\n## 🤖 Gateway Watchdog (v6.1) — Proactive Stability Layer\n\n**Overview:**\nA background daemon that periodically polls the Gateway health status. Runs independently from user requests, providing real"},{"path":"README.md","content":"# 🛡️ Autofix + Watchdog\n\n**OpenClaw diagnostic repair skill + Gateway health monitoring daemon.**\n\nWhen something breaks — Gateway won't start, tools return errors, configs go wrong — this skill walks through a structured cycle to diagnose, fix, and validate. The optional Watchdog daemon monitors Gateway health in the background and alerts you in real time.\n\n---\n\n## ✨ Features\n\n### Autofix Diagnosis (v6.1)\n- **Unified diagnosis pipeline** — 3 data sources in one pass (`gateway status` + runtime checks + key validation)\n- **`openclaw doctor` fallback** — 12s timeout, auto‑falls back to `gateway status --json` (fixes doctor hang in v2026.5.20)\n- **Safe subprocess cleanup** — All commands use `taskkill /T /F` to prevent zombie processes on Windows\n- **Interactive Health Dashboard** — HTML dashboard with Canvas embed (`health_dashboard.py --canvas`)\n- **Regression comparison** — `--save-baseline` / `--compare` to track fixes over time\n\n### Auto-Repair (v1.1)\n- **Gateway restart without hangs** — Uses `taskkill + start` instead of `openclaw gateway restart` (which hangs in v2026.5.20)\n- **Session archiving** — >7-day-old session files automatically archived to `archive/` when backlog >100\n- **Gateway memory alert** — Recommends restart when memory exceeds 800 MB\n- **Verification polling** — After repair, polls Gateway status every 5 seconds (up to 30s)\n\n### Gateway Watchdog *(optional)*\n- **Real-time health monitoring** — Polls Gateway every 60 seconds\n- **Dual-channel alerting** — WebChat session (async, ~40s) + Feishu DM (instant)\n- **Severity escalation** — Four-tier alerting (🟢/🟡/🟠/🔴) with noise filtering\n- **Auto-repair** — Low-risk issues (CLI path, session archive) are fixed automatically\n- **Stale process cleanup** — Auto-kills orphaned daemon processes on startup\n- **`--status` command** — Quick state check without starting a new daemon\n- **Single-instance protection** — Windows Mutex prevents duplicate daemons\n- **Auto-start on login** — Registers in `HKCU\\Run`\n\n---\n\n## 🚀 Quick Start\n\n### Just tell your agent:\n\n```\nrun autofix self-check\ncheck what's wrong with Gateway\nauto repair\n```\n\n### Recommended first step for any issue:\n\n```bash\npython scripts/diagnosis_formatter.py --json # Unified diagnosis\npython scripts/health_dashboard.py --canvas # Visual dashboard\n```\n\n> ⚠️ **v2026.5.20 note:** `openclaw doctor` may hang. The pipeline auto‑falls back to `gateway status --json` after 12 seconds.\n\n---\n\n## 📦 Installation\n\n```bash\nclawhub install autofix\n```\n\nThen restart OpenClaw to load the skill.\n\n### Watchdog daemon *(optional)*\n\n```bash\npip install -r scripts/requirements.txt\npython scripts/watchdog_monitor.py # Start daemon\npython scripts/watchdog_monitor.py --status # Check status\npython scripts/watchdog_monitor.py --install # Auto-start on login\n```\n\nSee [`INSTALL.md`](./INSTALL.md) for full setup.\n\n---\n\n## 🔔 Feishu Error Notification Setup\n\nWhen the Watchdog detects anomalies (Gateway c"},{"path":"_meta.json","content":"{\n \"ownerId\": \"kn78s99sqh4h99xeq9gm34c33582xjhg\",\n \"slug\": \"autofix\",\n \"version\": \"6.1.0\",\n \"publishedAt\": 1779445458720\n}"},{"path":"scripts/requirements.txt","content":"pywin32>=310\npsutil>=5.9.0"},{"path":"docs/COMPLETENESS_SUMMARY.md","content":"# ✅ autofix-theclaw v5.0 - 完整性验证摘要\r\n\r\n**生成时间：** 2026-05-17 21:30 \r\n**版本状态：** v4.5 → v5.0 (部分完成) \r\n**总体评分：** **75%** ⭐⭐⭐⭐ \r\n\r\n---\r\n\r\n## 📊 完整性概览\r\n\r\n| 组件 | 文件数 | 大小 | 状态 |\r\n|------|--------|------|------|\r\n| **核心文档** | 9 | 60KB | ✅ 100% 完整 |\r\n| **Canvas 脚本** | 2 | 13KB | ✅ 100% 完整 |\r\n| **Python 工具** | 2 | ❌ 待创建 | ⚠️ 40% 完成 |\r\n| **工作流集成** | - | - | ⚠️ 60% 集成 |\r\n\r\n---\r\n\r\n## ✅ 已完成的 v5.0 功能\r\n\r\n### **1. 诊断报告可视化 (DRE)** ✅ **可运行**\r\n- Canvas HTML 模板已实现（`CanvasScript_DiagnosticReport.js`, 9.2KB）\r\n- 风险标志、证据链对比图、回滚命令展示均已设计\r\n- Canvas 脚本语法正确，可直接嵌入回答\r\n\r\n### **2. 文档体系** ✅ **完整**\r\n- `MODULE_03_Enhancement_Reports.md`（10KB）：详细设计文档\r\n- `EXAMPLE_usage.md`（14KB）：使用示例和场景演示\r\n- `QUICK_START_v5.0.md`（10KB）：快速实施指南\r\n- `CHANGES_v5.0.md`（9KB）：完整变更总结\r\n\r\n### **3. SKILL.md 更新** ✅ **v5.0**\r\n- 主控文档已标注 v5.0\r\n- DRE 功能在模块索引中提及\r\n\r\n---\r\n\r\n## ⚠️ 待完成的工作（预计 20 分钟）\r\n\r\n### **Step 1: 创建 Python 工具函数** (10 分钟)\r\n\r\n#### A. ELIS 工具函数\r\n```python\r\n# 文件：tools/elis_helper.py\r\ndef analyze_error_logs(exec_output, problem_context):\r\n # TODO: 集成 LLM 调用逻辑\r\n return {\r\n \"core_issue\": \"\",\r\n \"causes\": [],\r\n \"fix_command\": \"\",\r\n \"risk_level\": \"Medium\",\r\n \"confidence_score\": 0.85\r\n }\r\n```\r\n\r\n#### B. Canvas 报告生成器\r\n```python\r\n# 文件：tools/canvas_report_generator.py\r\nclass CanvasReportGenerator:\r\n def generate_report(self, diag_data) -> str:\r\n # TODO: 封装 CanvasSnapshot 调用\r\n pass\r\n```\r\n\r\n### **Step 2: 更新 MODULE_03_ValidationAction.md** (10 分钟)\r\n\r\n在 Path B 部分添加 ELIS 和 Canvas 报告生成代码：\r\n```markdown\r\n### Path B: Code Verification → L1 Loop (v5.0 Enhanced)\r\n a. **[NEW] Extract and Analyze Error (ELIS)**\r\n ```python\r\n from autofix_theclaw.tools.elis_helper import analyze_error_logs\r\n analysis = analyze_error_logs(exec_result.output, problem_type)\r\n ```\r\n b. **[NEW] Generate Diagnosis Report (Canvas)**\r\n ```python\r\n from autofix_theclaw.tools.canvas_report_generator import CanvasReportGenerator\r\n generator = CanvasReportGenerator()\r\n report_html = generator.generate_report(analysis)\r\n ```\r\n```\r\n\r\n---\r\n\r\n## 🎯 最终完整性预期\r\n\r\n### **完成 Step 1-2 后：95%** ⭐⭐⭐⭐⭐\r\n\r\n| 评估维度 | 当前 | 预期 | 说明 |\r\n|---------|------|------|------|\r\n| 文档完整性 | 100% | 100% | ✅ 已完整 |\r\n| Canvas 脚本完整性 | 100% | 100% | ✅ 已完整 |\r\n| Python 工具完整性 | 40% | **95%** | ⚠️ → ✅ 待创建 |\r\n| 工作流集成 | 60% | **100%** | ⚠️ → ✅ 待更新 |\r\n\r\n---\r\n\r\n## 📚 详细验证报告\r\n\r\n完整验证报告见： \r\n`~\\.openclaw\\workspace\\skills\\autofix-theclaw\\VERIFICATION_REPORT.md`（7KB）\r\n\r\n该文件包含：\r\n- ✅ 目录结构验证清单\r\n- ⚠️ 功能完整性测试结果\r\n- 🚨 关键风险点分析\r\n- 🎯 下一步行动清单\r\n\r\n---\r\n\r\n## 🚀 快速部署命令\r\n\r\n```bash\r\n# Step 1: 查看验证报告\r\ntype ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\VERIFICATION_REPORT.md\r\n\r\n# Step 2: 创建 ELIS 工具函数（从 CHANGES_v5.0.md复制）\r\ncopy \"~\\.openclaw\\workspace\\skills\\autofix-theclaw\\CHANGES_v5.0.md\" .\r\n\r\n# Step 3: 运行 Canvas 脚本语法检查\r\npython -m py_compile ~\\.openclaw\\workspace\\skills\\autofix-theclaw\\tools\\CanvasScript_DiagnosticReport.js\r\n```\r\n\r\n---\r\n\r\n## 📞 需要帮助时联系\r\n\r\n- @autofix-team (内部团队)\r\n- OpenClaw Discord: `https://d"}],"languages":[],"docsSourceLabel":"CLAWHUB","editorialOverview":"A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates... Skill: OpenClaw Problem Solver自动修复小龙虾 Owner: mikewongonline Summary: A comprehensive, self-evolving skill designed to diagnose and solve OpenClaw issues by following a structured, multi-stage resolution cycle. It incorporates... Tags: latest:6.1.0 Version history: v6.1.0 | 2026-05-22T10:24:18.720Z | user **Gateway Watchdog reliability and notification pipeline improvements.** - Gateway Watchdog upgraded to v6.1 with","editorialQuality":{"score":100,"threshold":65,"status":"ready","wordCount":1629,"uniquenessScore":47,"reasons":[]}},"media":{"evidence":{"source":"no-media","verified":false,"confidence":"low","updatedAt":"2026-10-11T21:21:41.392Z","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":"2026-10-11T21:21:41.392Z","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":"2026-10-13T05:33:18.646Z","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"}]}}}