[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75M8SJS4A48NF3F16X13FJ9":3,"skill-same:01M75M8SJS4A48NF3F16X13FJ9":70},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":47,"commitSha":48,"skillDirectory":49,"forksCount":50,"repositoryHtmlUrl":51,"repositoryHomepageUrl":6,"ownerHtmlUrl":52,"ownerBlogUrl":53,"ownerDescription":54,"updatedAt":55,"install":56,"resources":69},"01M75M8SJS4A48NF3F16X13FJ9","safety-checks",null,"Verify before you trust — model pinning, fallbacks, and runtime safety validation","javimosch\u002Fopen-claw-skills","javimosch","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F12900508?v=4","01M6D3F2AS6PV75ET9PZ1QGP8J","Javier Leandro Arancibia","01M75B8NCTAJG9WH5YNZH7C3D8",0,"MIT","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fleegitw\u002Fsafety-checks\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: safety-checks\nversion: 1.5.2\ndescription: Verify before you trust — model pinning, fallbacks, and runtime safety validation\nauthor: Live Neon \u003Ccontact@liveneon.dev>\nhomepage: https:\u002F\u002Fgithub.com\u002Flive-neon\u002Fskills\u002Ftree\u002Fmain\u002Fagentic\u002Fsafety-checks\nrepository: leegitw\u002Fsafety-checks\nlicense: MIT\ntags: [agentic, safety, validation, model-pinning, fallback, runtime-verification, cache, guardrails]\nlayer: safety\nstatus: active\nalias: sc\nmetadata:\n  openclaw:\n    requires:\n      config:\n        - .openclaw\u002Fsafety-checks.yaml\n        - .claude\u002Fsafety-checks.yaml\n      workspace:\n        - output\u002Fsafety\u002F\n---\n\n# safety-checks (安全)\n\nUnified skill for runtime safety verification including model version pinning,\nfallback chain validation, cache staleness detection, and cross-session state checks.\nConsolidates 4 granular skills into a single safety verification suite.\n\n**Trigger**: 事前検証 (pre-flight) or HEARTBEAT\n\n**Source skills**: model-pinner, fallback-checker, cache-validator, cross-session-safety-check (from extensions)\n\n## Installation\n\n```bash\nopenclaw install leegitw\u002Fsafety-checks\n```\n\n**Dependencies**: `leegitw\u002Fconstraint-engine` (for enforcement integration)\n\n```bash\n# Install with dependencies\nopenclaw install leegitw\u002Fcontext-verifier\nopenclaw install leegitw\u002Ffailure-memory\nopenclaw install leegitw\u002Fconstraint-engine\nopenclaw install leegitw\u002Fsafety-checks\n```\n\n**Standalone usage**: Model pinning and cache checks work independently.\nFull integration with constraint enforcement requires constraint-engine.\n\n**Data handling**: This skill performs local-only operations. All checks (model version comparison,\ncache age verification, file lock detection) are local file\u002Fmetadata operations — no data is sent\nto any model, API, or external service. Results are written to `output\u002Fsafety\u002F` in your workspace.\n\n## What This Solves\n\nAI systems can silently degrade — model versions drift, caches go stale, sessions accumulate state. This skill catches these issues before they cause problems:\n\n1. **Model pinning** — verify you're using the model you expect\n2. **Fallback validation** — ensure degraded-mode paths exist and work\n3. **Cache checks** — detect stale or corrupted cached data\n4. **Session hygiene** — identify cross-session state contamination\n\n**The insight**: Runtime verification catches what static rules miss. Check the system state, not just the configuration.\n\n## Usage\n\n```\n\u002Fsc \u003Csub-command> [arguments]\n```\n\n## Sub-Commands\n\n| Command | CJK | Logic | Trigger |\n|---------|-----|-------|---------|\n| `\u002Fsc model` | 機種 | model.version→pinned✓∨drift✗ | HEARTBEAT |\n| `\u002Fsc fallback` | 代替 | chain.exists→safe✓∨missing✗ | HEARTBEAT |\n| `\u002Fsc cache` | 快取 | response.age>TTL→stale✗ | HEARTBEAT |\n| `\u002Fsc session` | 会話 | cross_session.state→clean✓∨interference✗ | HEARTBEAT |\n\n## Arguments\n\n### \u002Fsc model\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --expected | No | Expected model version (default: from config) |\n| --strict | No | Fail on any version mismatch |\n\n### \u002Fsc fallback\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --chain | No | Specific fallback chain to check in config |\n\n**Note**: This command validates that fallback configurations exist in your config file.\nIt does NOT make network calls or test actual connectivity. It's a config file audit.\n\n### \u002Fsc cache\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --ttl | No | TTL in seconds (default: 3600) |\n| --clear | No | Clear stale cache entries |\n\n### \u002Fsc session\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --check-state | No | Check for state leakage between sessions |\n| --clear-state | No | Clear any leaked state |\n\n## Configuration\n\nConfiguration is loaded from (in order of precedence):\n1. `.openclaw\u002Fsafety-checks.yaml` (OpenClaw standard)\n2. `.claude\u002Fsafety-checks.yaml` (Claude Code compatibility)\n3. Defaults (built-in)\n\n## Core Logic\n\n### Model Version Pinning\n\nEnsures AI model version matches expected configuration.\n\n**How it works**: This skill compares the model version reported in your agent's session\nmetadata (e.g., the model name in API responses or agent headers) against the expected\nversion in your config file. It does NOT call the model API to check — it reads the\nversion string that your agent runtime already exposes.\n\n**Model version format**: `{provider}-{model}-{version}-{date}`\n\nExamples:\n- `anthropic-opus-4-5-20251101`\n- `openai-gpt-4-turbo-20250301`\n- `google-gemini-2-pro-20260101`\n\n```yaml\n# .openclaw\u002Fsafety-checks.yaml\nmodel:\n  expected: \"anthropic-opus-4-5-20251101\"  # Provider-neutral format\n  strict: true\n```\n\n| Condition | Result |\n|-----------|--------|\n| Version matches | ✓ Pinned |\n| Version differs, strict=false | ⚠ Warning |\n| Version differs, strict=true | ✗ Fail |\n\n### Fallback Chain Validation\n\n**Config audit only** — verifies fallback entries exist in your config file. This does NOT\nmake network calls or test actual connectivity.\n\nChecks that your `safety-checks.yaml` declares fallback configurations:\n\n```yaml\n# Example config entries this command validates exist:\nfallbacks:\n  model: [\"primary-model\", \"fallback-model\", \"cached\"]\n  storage: [\"primary-path\", \"backup-path\"]\n```\n\n| Check | What It Validates |\n|-------|-------------------|\n| Model fallbacks | Config lists alternative models |\n| Storage fallbacks | Config lists backup paths |\n\n### Cache Staleness Detection\n\nPrevents use of outdated cached data:\n\n| Age | TTL | Status |\n|-----|-----|--------|\n| \u003C TTL | Any | ✓ Fresh |\n| > TTL | Not critical | ⚠ Stale warning |\n| > TTL | Critical | ✗ Stale fail |\n\n### Cross-Session State\n\nDetects state leakage between sessions:\n\n| Check | Exact File Checked | Risk |\n|-------|-------------------|------|\n| File locks | `.openclaw\u002Fsafety-checks.lock` | Resource contention |\n| Workspace temp files | `output\u002Fsafety\u002Ftemp-*` | Disk exhaustion |\n| Skill config | `.openclaw\u002Fsafety-checks.yaml` values | Configuration drift |\n\n**Scope**: Checks are limited to this skill's own files only. This skill does NOT scan\ndirectories, does NOT read other skills' config files, and does NOT access files outside\nthe specific paths listed above.\n\n## Output\n\n### \u002Fsc model output (pinned)\n\n```\n[MODEL CHECK]\nStatus: ✓ PINNED\n\nExpected: anthropic-opus-4-5-20251101\nActual:   anthropic-opus-4-5-20251101\n\nModel version matches configuration.\n```\n\n### \u002Fsc model output (drift)\n\n```\n[MODEL CHECK]\nStatus: ✗ VERSION DRIFT\n\nExpected: anthropic-opus-4-5-20251101\nActual:   anthropic-opus-4-5-20251201\n\nWARNING: Model version has changed.\nThis may affect behavior consistency.\n\nAction: Update expected version in settings, or investigate change.\n```\n\n### \u002Fsc fallback output\n\n```\n[FALLBACK CHECK]\nStatus: ✓ CONFIGURED\n\nConfig file: .openclaw\u002Fsafety-checks.yaml\n\nFallback entries found:\n  ✓ model.fallbacks: 3 entries defined\n  ✓ storage.fallbacks: 2 entries defined\n\nNote: This validates config entries exist, not actual connectivity.\n  ✓ Tertiary (memory) - 4GB free\n```\n\n### \u002Fsc cache output\n\n```\n[CACHE CHECK]\nStatus: ⚠ STALE ENTRIES FOUND\n\nCache Statistics:\n  Total entries: 156\n  Fresh (\u003C 1h): 142\n  Stale (> 1h): 14\n  Critical stale: 0\n\nStale entries:\n  - api_response_auth (age: 2h 15m)\n  - user_preferences (age: 1h 30m)\n  - ...\n\nAction: Run \u002Fsc cache --clear to remove stale entries.\n```\n\n### \u002Fsc session output (clean)\n\n```\n[SESSION CHECK]\nStatus: ✓ CLEAN\n\nNo cross-session interference detected.\n\nChecks passed:\n  ✓ No stale lock file (.openclaw\u002Fsafety-checks.lock)\n  ✓ No orphan temp files (output\u002Fsafety\u002Ftemp-*)\n  ✓ Config file valid (.openclaw\u002Fsafety-checks.yaml)\n```\n\n### \u002Fsc session output (interference)\n\n```\n[SESSION CHECK]\nStatus: ✗ INTERFERENCE DETECTED\n\nIssues found:\n\n1. Stale file lock:\n   File: .openclaw\u002Fsafety-checks.lock\n   Owner: PID 12345 (not running)\n   Action: Remove lock with \u002Fsc session --clear-state\n\n2. Orphan temp files (3):\n   output\u002Fsafety\u002Ftemp-*.log\n   Age: > 24 hours\n   Action: Remove with \u002Fsc session --clear-state\n\nRun \u002Fsc session --clear-state to remediate.\n```\n\n## Integration\n\n- **Layer**: Safety\n- **Depends on**: constraint-engine (for enforcement integration)\n- **Used by**: constraint-engine (for pre-action safety checks), governance (for health monitoring)\n\n## Failure Modes\n\n| Condition | Behavior |\n|-----------|----------|\n| Invalid sub-command | List available sub-commands |\n| Config not found | Use defaults, warn user |\n| Network unavailable | Skip remote checks, warn |\n| Permission denied | Error with remediation steps |\n\n## Next Steps\n\nAfter invoking this skill:\n\n| Condition | Action |\n|-----------|--------|\n| Model drift | Alert user, suggest config update |\n| Fallback missing | Alert user, suggest setup |\n| Cache stale | Offer to clear, or auto-clear if configured |\n| Session interference | Clear state, log incident |\n\n## Workspace Files\n\nThis skill reads\u002Fwrites only within declared paths:\n\n```\n.openclaw\u002F\n├── safety-checks.yaml     # Primary config (read)\n└── cache\u002F\n    └── staleness.log      # Cache check history (read\u002Fwrite)\n\n.claude\u002F\n└── safety-checks.yaml     # Claude Code config (read)\n\noutput\u002F\n└── safety\u002F\n    ├── model-checks.log   # Model version history (write)\n    ├── fallback-tests.log # Fallback test results (write)\n    └── session-state.log  # Session state log (write)\n```\n\n**Note**: This skill does NOT read `.claude\u002Fsettings.json` or any other tool's configuration.\n\n## HEARTBEAT Integration\n\nThese checks should run periodically via HEARTBEAT:\n\n```markdown\n## P1: Critical (Every Session)\n- [ ] Model version pinned? → \u002Fsc model\n- [ ] Session state clean? → \u002Fsc session\n\n## P2: Important (Weekly)\n- [ ] Fallback chains healthy? → \u002Fsc fallback\n- [ ] Cache fresh? → \u002Fsc cache\n```\n\n## Examples\n\n### Model Drift Detection\n\n```\n\u002Fsc model --strict\n[MODEL CHECK]\nStatus: ⚠ VERSION DRIFT\n\nExpected: anthropic-opus-4-5-20251101\nActual:   anthropic-opus-4-5-20260101\n\nModel version changed. Review CHANGELOG for behavior changes.\n```\n\n### Cache Cleanup\n\n```\n\u002Fsc cache --clear\n[CACHE CHECK]\nStatus: ✓ CLEANED\n\nRemoved 14 stale entries from .openclaw\u002Fcache\u002F\nFreed: 2.3 MB\n```\n\n## Security Considerations\n\n**Local-only processing**: All safety checks are local file and metadata operations. No data is\nsent to any LLM, API, or external service. The \"agent's model\" is only used to interpret your\ncommands — not to process or transmit your data.\n\n**Exact files accessed (read):**\n- `.openclaw\u002Fsafety-checks.yaml` — your skill configuration\n- `.claude\u002Fsafety-checks.yaml` — alternate config location\n- `.openclaw\u002Fcache\u002Fstaleness.log` — cache check history (if exists)\n\n**Exact files written:**\n- `output\u002Fsafety\u002F*.log` — check results and history\n\nThis skill does NOT scan directories or read other skills' config files.\n\n**What this skill does NOT access:**\n- Other skills' configuration files\n- System environment variables\n- `.claude\u002Fsettings.json` or other tool configs\n- Any files outside the three config paths listed above\n- Network resources or external APIs\n\n**What this skill does NOT do:**\n- Send data to any model or external service\n- Modify files outside `output\u002Fsafety\u002F`\n- Execute arbitrary code\n\n**Data handling:**\n- All checks are read-only except for `--clear` operations on its own cache\u002Foutput\n- Results are written to `output\u002Fsafety\u002F` only\n- No data leaves the local machine\n\n**Model version clarification:**\nThe `\u002Fsc model` command compares the expected version in your config against the model\nversion string reported by your agent runtime (e.g., in session metadata or API headers).\nIt does NOT make API calls to verify the model — it reads information your agent already has.\n\n**Provenance note:**\nThis skill is developed by Live Neon (https:\u002F\u002Fgithub.com\u002Flive-neon\u002Fskills) and published\nto ClawHub under the `leegitw` account. Both refer to the same maintainer.\n\n## Acceptance Criteria\n\n- [ ] `\u002Fsc model` verifies model version matches config\n- [ ] `\u002Fsc model` warns or fails on version drift based on strict setting\n- [ ] `\u002Fsc fallback` verifies fallback config entries exist in config file\n- [ ] `\u002Fsc cache` detects entries older than TTL\n- [ ] `\u002Fsc cache --clear` removes stale entries\n- [ ] `\u002Fsc session` detects cross-session state leakage\n- [ ] `\u002Fsc session --clear-state` remediates found issues\n- [ ] All checks integrated with HEARTBEAT\n\n---\n\n*Consolidated from 4 skills as part of agentic skills consolidation (2026-02-15).*\n","# safety-checks (安全)\n\nUnified skill for runtime safety verification including model version pinning,\nfallback chain validation, cache staleness detection, and cross-session state checks.\nConsolidates 4 granular skills into a single safety verification suite.\n\n**Trigger**: 事前検証 (pre-flight) or HEARTBEAT\n\n**Source skills**: model-pinner, fallback-checker, cache-validator, cross-session-safety-check (from extensions)\n\n## Installation\n\n```bash\nopenclaw install leegitw\u002Fsafety-checks\n```\n\n**Dependencies**: `leegitw\u002Fconstraint-engine` (for enforcement integration)\n\n```bash\n# Install with dependencies\nopenclaw install leegitw\u002Fcontext-verifier\nopenclaw install leegitw\u002Ffailure-memory\nopenclaw install leegitw\u002Fconstraint-engine\nopenclaw install leegitw\u002Fsafety-checks\n```\n\n**Standalone usage**: Model pinning and cache checks work independently.\nFull integration with constraint enforcement requires constraint-engine.\n\n**Data handling**: This skill performs local-only operations. All checks (model version comparison,\ncache age verification, file lock detection) are local file\u002Fmetadata operations — no data is sent\nto any model, API, or external service. Results are written to `output\u002Fsafety\u002F` in your workspace.\n\n## What This Solves\n\nAI systems can silently degrade — model versions drift, caches go stale, sessions accumulate state. This skill catches these issues before they cause problems:\n\n1. **Model pinning** — verify you're using the model you expect\n2. **Fallback validation** — ensure degraded-mode paths exist and work\n3. **Cache checks** — detect stale or corrupted cached data\n4. **Session hygiene** — identify cross-session state contamination\n\n**The insight**: Runtime verification catches what static rules miss. Check the system state, not just the configuration.\n\n## Usage\n\n```\n\u002Fsc \u003Csub-command> [arguments]\n```\n\n## Sub-Commands\n\n| Command | CJK | Logic | Trigger |\n|---------|-----|-------|---------|\n| `\u002Fsc model` | 機種 | model.version→pinned✓∨drift✗ | HEARTBEAT |\n| `\u002Fsc fallback` | 代替 | chain.exists→safe✓∨missing✗ | HEARTBEAT |\n| `\u002Fsc cache` | 快取 | response.age>TTL→stale✗ | HEARTBEAT |\n| `\u002Fsc session` | 会話 | cross_session.state→clean✓∨interference✗ | HEARTBEAT |\n\n## Arguments\n\n### \u002Fsc model\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --expected | No | Expected model version (default: from config) |\n| --strict | No | Fail on any version mismatch |\n\n### \u002Fsc fallback\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --chain | No | Specific fallback chain to check in config |\n\n**Note**: This command validates that fallback configurations exist in your config file.\nIt does NOT make network calls or test actual connectivity. It's a config file audit.\n\n### \u002Fsc cache\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --ttl | No | TTL in seconds (default: 3600) |\n| --clear | No | Clear stale cache entries |\n\n### \u002Fsc session\n\n| Argument | Required | Description |\n|----------|----------|-------------|\n| --check-state | No | Check for state leakage between sessions |\n| --clear-state | No | Clear any leaked state |\n\n## Configuration\n\nConfiguration is loaded from (in order of precedence):\n1. `.openclaw\u002Fsafety-checks.yaml` (OpenClaw standard)\n2. `.claude\u002Fsafety-checks.yaml` (Claude Code compatibility)\n3. Defaults (built-in)\n\n## Core Logic\n\n### Model Version Pinning\n\nEnsures AI model version matches expected configuration.\n\n**How it works**: This skill compares the model version reported in your agent's session\nmetadata (e.g., the model name in API responses or agent headers) against the expected\nversion in your config file. It does NOT call the model API to check — it reads the\nversion string that your agent runtime already exposes.\n\n**Model version format**: `{provider}-{model}-{version}-{date}`\n\nExamples:\n- `anthropic-opus-4-5-20251101`\n- `openai-gpt-4-turbo-20250301`\n- `google-gemini-2-pro-20260101`\n\n```yaml\n# .openclaw\u002Fsafety-checks.yaml\nmodel:\n  expected: \"anthropic-opus-4-5-20251101\"  # Provider-neutral format\n  strict: true\n```\n\n| Condition | Result |\n|-----------|--------|\n| Version matches | ✓ Pinned |\n| Version differs, strict=false | ⚠ Warning |\n| Version differs, strict=true | ✗ Fail |\n\n### Fallback Chain Validation\n\n**Config audit only** — verifies fallback entries exist in your config file. This does NOT\nmake network calls or test actual connectivity.\n\nChecks that your `safety-checks.yaml` declares fallback configurations:\n\n```yaml\n# Example config entries this command validates exist:\nfallbacks:\n  model: [\"primary-model\", \"fallback-model\", \"cached\"]\n  storage: [\"primary-path\", \"backup-path\"]\n```\n\n| Check | What It Validates |\n|-------|-------------------|\n| Model fallbacks | Config lists alternative models |\n| Storage fallbacks | Config lists backup paths |\n\n### Cache Staleness Detection\n\nPrevents use of outdated cached data:\n\n| Age | TTL | Status |\n|-----|-----|--------|\n| \u003C TTL | Any | ✓ Fresh |\n| > TTL | Not critical | ⚠ Stale warning |\n| > TTL | Critical | ✗ Stale fail |\n\n### Cross-Session State\n\nDetects state leakage between sessions:\n\n| Check | Exact File Checked | Risk |\n|-------|-------------------|------|\n| File locks | `.openclaw\u002Fsafety-checks.lock` | Resource contention |\n| Workspace temp files | `output\u002Fsafety\u002Ftemp-*` | Disk exhaustion |\n| Skill config | `.openclaw\u002Fsafety-checks.yaml` values | Configuration drift |\n\n**Scope**: Checks are limited to this skill's own files only. This skill does NOT scan\ndirectories, does NOT read other skills' config files, and does NOT access files outside\nthe specific paths listed above.\n\n## Output\n\n### \u002Fsc model output (pinned)\n\n```\n[MODEL CHECK]\nStatus: ✓ PINNED\n\nExpected: anthropic-opus-4-5-20251101\nActual:   anthropic-opus-4-5-20251101\n\nModel version matches configuration.\n```\n\n### \u002Fsc model output (drift)\n\n```\n[MODEL CHECK]\nStatus: ✗ VERSION DRIFT\n\nExpected: anthropic-opus-4-5-20251101\nActual:   anthropic-opus-4-5-20251201\n\nWARNING: Model version has changed.\nThis may affect behavior consistency.\n\nAction: Update expected version in settings, or investigate change.\n```\n\n### \u002Fsc fallback output\n\n```\n[FALLBACK CHECK]\nStatus: ✓ CONFIGURED\n\nConfig file: .openclaw\u002Fsafety-checks.yaml\n\nFallback entries found:\n  ✓ model.fallbacks: 3 entries defined\n  ✓ storage.fallbacks: 2 entries defined\n\nNote: This validates config entries exist, not actual connectivity.\n  ✓ Tertiary (memory) - 4GB free\n```\n\n### \u002Fsc cache output\n\n```\n[CACHE CHECK]\nStatus: ⚠ STALE ENTRIES FOUND\n\nCache Statistics:\n  Total entries: 156\n  Fresh (\u003C 1h): 142\n  Stale (> 1h): 14\n  Critical stale: 0\n\nStale entries:\n  - api_response_auth (age: 2h 15m)\n  - user_preferences (age: 1h 30m)\n  - ...\n\nAction: Run \u002Fsc cache --clear to remove stale entries.\n```\n\n### \u002Fsc session output (clean)\n\n```\n[SESSION CHECK]\nStatus: ✓ CLEAN\n\nNo cross-session interference detected.\n\nChecks passed:\n  ✓ No stale lock file (.openclaw\u002Fsafety-checks.lock)\n  ✓ No orphan temp files (output\u002Fsafety\u002Ftemp-*)\n  ✓ Config file valid (.openclaw\u002Fsafety-checks.yaml)\n```\n\n### \u002Fsc session output (interference)\n\n```\n[SESSION CHECK]\nStatus: ✗ INTERFERENCE DETECTED\n\nIssues found:\n\n1. Stale file lock:\n   File: .openclaw\u002Fsafety-checks.lock\n   Owner: PID 12345 (not running)\n   Action: Remove lock with \u002Fsc session --clear-state\n\n2. Orphan temp files (3):\n   output\u002Fsafety\u002Ftemp-*.log\n   Age: > 24 hours\n   Action: Remove with \u002Fsc session --clear-state\n\nRun \u002Fsc session --clear-state to remediate.\n```\n\n## Integration\n\n- **Layer**: Safety\n- **Depends on**: constraint-engine (for enforcement integration)\n- **Used by**: constraint-engine (for pre-action safety checks), governance (for health monitoring)\n\n## Failure Modes\n\n| Condition | Behavior |\n|-----------|----------|\n| Invalid sub-command | List available sub-commands |\n| Config not found | Use defaults, warn user |\n| Network unavailable | Skip remote checks, warn |\n| Permission denied | Error with remediation steps |\n\n## Next Steps\n\nAfter invoking this skill:\n\n| Condition | Action |\n|-----------|--------|\n| Model drift | Alert user, suggest config update |\n| Fallback missing | Alert user, suggest setup |\n| Cache stale | Offer to clear, or auto-clear if configured |\n| Session interference | Clear state, log incident |\n\n## Workspace Files\n\nThis skill reads\u002Fwrites only within declared paths:\n\n```\n.openclaw\u002F\n├── safety-checks.yaml     # Primary config (read)\n└── cache\u002F\n    └── staleness.log      # Cache check history (read\u002Fwrite)\n\n.claude\u002F\n└── safety-checks.yaml     # Claude Code config (read)\n\noutput\u002F\n└── safety\u002F\n    ├── model-checks.log   # Model version history (write)\n    ├── fallback-tests.log # Fallback test results (write)\n    └── session-state.log  # Session state log (write)\n```\n\n**Note**: This skill does NOT read `.claude\u002Fsettings.json` or any other tool's configuration.\n\n## HEARTBEAT Integration\n\nThese checks should run periodically via HEARTBEAT:\n\n```markdown\n## P1: Critical (Every Session)\n- [ ] Model version pinned? → \u002Fsc model\n- [ ] Session state clean? → \u002Fsc session\n\n## P2: Important (Weekly)\n- [ ] Fallback chains healthy? → \u002Fsc fallback\n- [ ] Cache fresh? → \u002Fsc cache\n```\n\n## Examples\n\n### Model Drift Detection\n\n```\n\u002Fsc model --strict\n[MODEL CHECK]\nStatus: ⚠ VERSION DRIFT\n\nExpected: anthropic-opus-4-5-20251101\nActual:   anthropic-opus-4-5-20260101\n\nModel version changed. Review CHANGELOG for behavior changes.\n```\n\n### Cache Cleanup\n\n```\n\u002Fsc cache --clear\n[CACHE CHECK]\nStatus: ✓ CLEANED\n\nRemoved 14 stale entries from .openclaw\u002Fcache\u002F\nFreed: 2.3 MB\n```\n\n## Security Considerations\n\n**Local-only processing**: All safety checks are local file and metadata operations. No data is\nsent to any LLM, API, or external service. The \"agent's model\" is only used to interpret your\ncommands — not to process or transmit your data.\n\n**Exact files accessed (read):**\n- `.openclaw\u002Fsafety-checks.yaml` — your skill configuration\n- `.claude\u002Fsafety-checks.yaml` — alternate config location\n- `.openclaw\u002Fcache\u002Fstaleness.log` — cache check history (if exists)\n\n**Exact files written:**\n- `output\u002Fsafety\u002F*.log` — check results and history\n\nThis skill does NOT scan directories or read other skills' config files.\n\n**What this skill does NOT access:**\n- Other skills' configuration files\n- System environment variables\n- `.claude\u002Fsettings.json` or other tool configs\n- Any files outside the three config paths listed above\n- Network resources or external APIs\n\n**What this skill does NOT do:**\n- Send data to any model or external service\n- Modify files outside `output\u002Fsafety\u002F`\n- Execute arbitrary code\n\n**Data handling:**\n- All checks are read-only except for `--clear` operations on its own cache\u002Foutput\n- Results are written to `output\u002Fsafety\u002F` only\n- No data leaves the local machine\n\n**Model version clarification:**\nThe `\u002Fsc model` command compares the expected version in your config against the model\nversion string reported by your agent runtime (e.g., in session metadata or API headers).\nIt does NOT make API calls to verify the model — it reads information your agent already has.\n\n**Provenance note:**\nThis skill is developed by Live Neon (https:\u002F\u002Fgithub.com\u002Flive-neon\u002Fskills) and published\nto ClawHub under the `leegitw` account. Both refer to the same maintainer.\n\n## Acceptance Criteria\n\n- [ ] `\u002Fsc model` verifies model version matches config\n- [ ] `\u002Fsc model` warns or fails on version drift based on strict setting\n- [ ] `\u002Fsc fallback` verifies fallback config entries exist in config file\n- [ ] `\u002Fsc cache` detects entries older than TTL\n- [ ] `\u002Fsc cache --clear` removes stale entries\n- [ ] `\u002Fsc session` detects cross-session state leakage\n- [ ] `\u002Fsc session --clear-state` remediates found issues\n- [ ] All checks integrated with HEARTBEAT\n\n---\n\n*Consolidated from 4 skills as part of agentic skills consolidation (2026-02-15).*\n",{"name":5,"tags":24,"alias":33,"layer":26,"author":34,"status":35,"license":15,"version":36,"homepage":37,"metadata":38,"repository":46,"description":7},[25,26,27,28,29,30,31,32],"agentic","safety","validation","model-pinning","fallback","runtime-verification","cache","guardrails","sc","Live Neon \u003Ccontact@liveneon.dev>","active","1.5.2","https:\u002F\u002Fgithub.com\u002Flive-neon\u002Fskills\u002Ftree\u002Fmain\u002Fagentic\u002Fsafety-checks",{"openclaw":39},{"requires":40},{"config":41,"workspace":44},[42,43],".openclaw\u002Fsafety-checks.yaml",".claude\u002Fsafety-checks.yaml",[45],"output\u002Fsafety\u002F","leegitw\u002Fsafety-checks","main","f9d3fe7cf9ac8f1ea63a0d96e98e1f510a3814db","skills\u002Fleegitw\u002Fsafety-checks",1,"https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills","https:\u002F\u002Fgithub.com\u002Fjavimosch","intrane.fr","Building robust, open-source infrastructure for the commons. AI-leveraged, self-hosted, built to last.","2026-06-07 03:34:58",{"source":8,"skillName":5,"commands":57,"prompt":67,"zipAvailable":68},[58,61,64],{"tool":59,"command":60},"npx","npx skills add javimosch\u002Fopen-claw-skills --skill safety-checks",{"tool":62,"command":63},"bunx","bunx skills add javimosch\u002Fopen-claw-skills --skill safety-checks",{"tool":65,"command":66},"pnpm","pnpm dlx skills add javimosch\u002Fopen-claw-skills --skill safety-checks","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fleegitw\u002Fsafety-checks\u002FSKILL.md",true,[],[71,79,86,93,100,107],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":78},"01M778GGXG2VQ0YF31W2MKGMQZ","crowd-prompting","A marketplace where AI agents improve prompts, system instructions, tool descriptions, and other text-based content with domain expertise from real-world operations — and earn tokens for valuable contributions.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzzadrian\u002Fcrowd-prompting\u002FSKILL.md",[],[],"2026-11-15 14:19:17",{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":78},"01M778G7BH8SBJ9FJ6C3G7CB1K","ceaser","Interact with the Ceaser privacy protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- shield, unshield, note management, and protocol queries. All ceaser tool calls use CLI subcommands (npx -y ceaser-mcp \u003Csubcommand>). Ceaser is a privacy-preserving ETH wrapper using Noir\u002FUltraHonk zero-knowledge proofs on Base L2 (chain ID 8453).","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyra-v21\u002Fceaser\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":78},"01M778FYCJQZWFKKX5FYMK3S47","ceaser-send","Fully automated private ETH transfer via Ceaser Protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- all ceaser tool calls use CLI subcommands (npx -y ceaser-mcp \u003Csubcommand>). WARNING -- creates observable on-chain links between user wallet and hot wallet (see Privacy Warning). Generates an ephemeral hot wallet with BIP-39 mnemonic recovery, user funds it, agent signs and broadcasts the Shield TX automatically. Exactly one manual step required (funding the hot wallet with ETH). Uses Noir\u002FUltraHonk zero-knowledge proofs.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyra-v21\u002Fceaser-send\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":78},"01M778FQQNSEYA3HJKZBPQ4M9Z","dilemm-ai","Compete in DilemmAI, the prisoner's dilemma AI arena at dilemm.ai. Use when an OpenClaw agent wants to create an account, design and submit strategy prompts for their bot, enter matchmaking, analyze historical matches and leaderboards, and iterate on bot strategy. Requires headless browser automation (no API available yet).","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyori\u002Fdilemm-ai\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":78},"01M778FGVZX7QA2TF9BFMM37FY","moldium","Post and manage content on the Moldium blog platform. Triggered by \"post to Moldium\", \"write a blog post\", \"publish an article\", etc.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyom45\u002Fmoldium\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":78},"01M778F88D8PQ9MQ0KNW2TJ8ZH","arxiv-paper-reviews","Interact with arXiv Crawler API to fetch papers, read reviews, submit comments, search papers, and import papers. Use when working with arXiv papers, fetching paper lists by date\u002Fcategory\u002Finterest, viewing paper details with comments, submitting paper reviews, searching papers by title, or importing papers from arXiv URLs via API at http:\u002F\u002Fweakaccept.top:8000\u002F.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzxrys\u002Farxiv-paper-reviews\u002FSKILL.md",[],[]]