openclaw-signet-pro

Review·Scanned 2/17/2026

This skill provides local integrity management for an OpenClaw workspace with commands like sign, verify, protect, quarantine, snapshot, and restore. It includes executable hooks that run python3 scripts/signet.py (e.g., a SessionStart hook) and installation instructions using git clone https://github.com/AtlasPA/openclaw-signet-pro.git, and reads OPENCLAW_WORKSPACE.

from clawhub.ai·v71d28c1·33.4 KB·0 installs
Scanned from 1.0.0 at 71d28c1 · Transparency log ↗
$ vett add clawhub.ai/atlaspa/openclaw-signet-proReview findings below

OpenClaw Signet Pro

Full cryptographic skill verification suite for OpenClaw, Claude Code, and any Agent Skills-compatible tool.

Everything in openclaw-signet (free) plus automated countermeasures: reject unsigned skills, quarantine tampered ones, create trusted snapshots, and restore from them.

Free Version Alerts. Pro Version Responds.

FeatureFreePro
SHA-256 skill signingYesYes
Tamper detectionYesYes
File-level change reportYesYes
Trust manifestYesYes
Quick status checkYesYes
Reject unsigned skills-Yes
Quarantine tampered skills-Yes
Tampering evidence collection-Yes
Trusted snapshots-Yes
Restore from snapshot-Yes
Automated protection sweep-Yes

Install

# Clone
git clone https://github.com/AtlasPA/openclaw-signet-pro.git

# Copy to your workspace skills directory
cp -r openclaw-signet-pro ~/.openclaw/workspace/skills/

Usage

# Sign all installed skills
python3 scripts/signet.py sign

# Sign a specific skill
python3 scripts/signet.py sign openclaw-warden

# Verify all skills
python3 scripts/signet.py verify

# List signed skills
python3 scripts/signet.py list

# Quick status
python3 scripts/signet.py status

# FULL PROTECTION SWEEP (recommended for session startup)
python3 scripts/signet.py protect

# Strict mode: also reject unsigned skills
python3 scripts/signet.py protect --reject-unsigned

# Reject unsigned skills
python3 scripts/signet.py reject

# Quarantine a tampered skill
python3 scripts/signet.py quarantine bad-skill

# Unquarantine after investigation
python3 scripts/signet.py unquarantine bad-skill

# Create trusted snapshot of a verified skill
python3 scripts/signet.py snapshot openclaw-warden

# Restore a skill from its trusted snapshot
python3 scripts/signet.py restore openclaw-warden

All commands accept --workspace /path/to/workspace. If omitted, auto-detects from $OPENCLAW_WORKSPACE, current directory, or ~/.openclaw/workspace.

How It Works

sign:        Compute SHA-256 of every file -> composite hash -> store in manifest
verify:      Recompute hashes -> compare to manifest -> report differences
reject:      Move unsigned skills to .quarantine/signet/ with metadata
quarantine:  Rename tampered skill with .quarantined- prefix, save evidence
snapshot:    Copy verified skill to .signet/snapshots/ with integrity metadata
restore:     Verify snapshot integrity -> replace skill dir -> update manifest
protect:     verify + quarantine tampered + (optionally reject unsigned) + snapshot clean

Countermeasures

CommandAction
protectFull sweep: verify all, quarantine tampered, snapshot clean
protect --reject-unsignedFull sweep + reject unsigned skills
reject [skill]Move unsigned skills to quarantine area
quarantine <skill>Disable tampered skill with evidence recording
unquarantine <skill>Re-enable quarantined skill (re-sign recommended)
snapshot <skill>Create trusted backup of a verified skill
restore <skill>Restore from trusted snapshot with integrity check

Session Startup Integration

Claude Code (hooks)

Add to .claude/settings.json:

{
  "hooks": {
    "SessionStart": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "python3 scripts/signet.py protect",
            "timeout": 30
          }
        ]
      }
    ]
  }
}

OpenClaw (heartbeat)

Add to HEARTBEAT.md:

- Run skill signature protection

Requirements

  • Python 3.8+
  • No external dependencies (stdlib only)
  • Cross-platform: Windows, macOS, Linux

License

MIT