{"skill":{"slug":"openclaw-warden","displayName":"Openclaw Warden","summary":"Verify workspace file integrity and scan for prompt injection patterns in agent identity and memory files. Detects unauthorized modifications to SOUL.md, AGENTS.md, IDENTITY.md, memory files, and installed skills. Free detection layer — upgrade to openclaw-warden-pro for automated countermeasures.","description":"---\nname: openclaw-warden\nuser-invocable: true\nmetadata: {\"openclaw\":{\"emoji\":\"🛡️\",\"requires\":{\"bins\":[\"python3\"]},\"os\":[\"darwin\",\"linux\",\"win32\"]}}\n---\n\n# OpenClaw Warden\n\nMonitors your workspace files for unauthorized modifications and prompt injection attacks. Existing security tools scan *skills* before installation — this tool watches the *workspace itself* after installation, catching tampering that other tools miss.\n\n## Why This Matters\n\nYour agent reads SOUL.md, AGENTS.md, IDENTITY.md, USER.md, and memory files on every session startup and **trusts them implicitly**. A compromised skill, a malicious heartbeat payload, or an unauthorized process can modify these files to:\n\n- Inject hidden instructions that alter agent behavior\n- Embed data exfiltration URLs in markdown images\n- Override identity and safety boundaries\n- Plant persistent backdoors in memory files\n\nThis skill detects all of these.\n\n\n## Commands\n\n### Establish Baseline\n\nCreate or reset the integrity baseline. Run this after setting up your workspace or after reviewing and accepting all current file states.\n\n```bash\npython3 {baseDir}/scripts/integrity.py baseline --workspace /path/to/workspace\n```\n\n### Verify Integrity\n\nCheck all monitored files against the stored baseline. Reports modifications, deletions, and new untracked files.\n\n```bash\npython3 {baseDir}/scripts/integrity.py verify --workspace /path/to/workspace\n```\n\n### Scan for Injections\n\nScan workspace files for prompt injection patterns: hidden instructions, base64 payloads, Unicode tricks, markdown image exfiltration, HTML injection, and suspicious system prompt markers.\n\n```bash\npython3 {baseDir}/scripts/integrity.py scan --workspace /path/to/workspace\n```\n\n### Full Check (Verify + Scan)\n\nRun both integrity verification and injection scanning in one pass.\n\n```bash\npython3 {baseDir}/scripts/integrity.py full --workspace /path/to/workspace\n```\n\n### Quick Status\n\nOne-line summary of workspace health.\n\n```bash\npython3 {baseDir}/scripts/integrity.py status --workspace /path/to/workspace\n```\n\n### Accept Changes\n\nAfter reviewing a legitimate change, update the baseline for a specific file.\n\n```bash\npython3 {baseDir}/scripts/integrity.py accept SOUL.md --workspace /path/to/workspace\n```\n\n## Workspace Auto-Detection\n\nIf `--workspace` is omitted, the script tries:\n1. `OPENCLAW_WORKSPACE` environment variable\n2. Current directory (if AGENTS.md exists)\n3. `~/.openclaw/workspace` (default)\n\n## What Gets Monitored\n\n| Category | Files | Alert Level on Change |\n|----------|-------|-----------------------|\n| **Critical** | SOUL.md, AGENTS.md, IDENTITY.md, USER.md, TOOLS.md, HEARTBEAT.md | WARNING |\n| **Memory** | memory/*.md, MEMORY.md | INFO (expected to change) |\n| **Config** | *.json in workspace root | WARNING |\n| **Skills** | skills/*/SKILL.md | WARNING |\n\nInjection patterns trigger **CRITICAL** alerts regardless of file category.\n\n## Injection Patterns Detected\n\n- **Instruction override:** \"ignore previous instructions\", \"disregard above\", \"you are now\", \"new system prompt\"\n- **Base64 payloads:** Suspiciously long base64 strings outside code blocks\n- **Unicode manipulation:** Zero-width characters, RTL overrides, homoglyphs\n- **Markdown exfiltration:** Image tags with data-encoding URLs\n- **HTML injection:** script tags, iframes, hidden elements\n- **System prompt markers:** `<system>`, `[SYSTEM]`, `<<SYS>>` blocks\n- **Shell injection:** `$(...)` outside code blocks\n\n## Exit Codes\n\n- `0` — Clean, no issues\n- `1` — Modifications detected (review needed)\n- `2` — Injection patterns detected (action needed)\n\n## No External Dependencies\n\nPython standard library only. No pip install. No network calls. Everything runs locally.\n\n## Cross-Platform\n\nWorks with OpenClaw, Claude Code, Cursor, and any tool using the Agent Skills specification.\n","tags":{"latest":"1.0.3"},"stats":{"comments":0,"downloads":2296,"installsAllTime":86,"installsCurrent":5,"stars":1,"versions":6},"createdAt":1770276296458,"updatedAt":1778486025837},"latestVersion":{"version":"1.0.3","createdAt":1770892012579,"changelog":"openclaw-warden 1.0.3\n\n- Removed promotional references to openclaw-warden-pro from documentation.\n- Cleaned up SKILL.md metadata and description formatting for clarity.\n- No functional code changes; this update is documentation-only.","license":null},"metadata":{"setup":[],"os":["darwin","linux","win32"],"systems":null},"owner":{"handle":"atlaspa","userId":"s17cbdkg4hkk4m581rvtwd7t5x885p6e","displayName":"AtlasPA","image":"https://avatars.githubusercontent.com/u/231540010?v=4"},"moderation":{"isSuspicious":false,"isMalwareBlocked":false,"verdict":"clean","reasonCodes":["review.llm_review"],"summary":"Review: review.llm_review","engineVersion":"v2.4.24","updatedAt":1779949163848}}