Claw Guard Server Skill

简体中文

This skill helps agents perform secure skill-hash verification and guard/report API calls with raw HTTP.

What This Skill Can Do

• Compute skill_scan_artifacts.manifestHash locally with a deterministic manifest-hash algorithm.
• Query POST /v1/guard/skill-hash to check server-side status for a known manifest hash.
• Run pre-action policy checks with POST /v1/guard/evaluate (allow, warn, block).
• Record post-action audit logs with POST /v1/guard/events.
• Report hash or zip samples with:
  • POST /v1/skill-report/upload-hash
  • POST /v1/skill-report/upload-file
• Use raw HTTP only, with examples for both curl (macOS/Linux) and PowerShell (Windows).

Security Notes

• This skill is for verification and evidence collection, not automatic trust.
• If a hash is unknown or response signals ambiguity, require manual confirmation.
• upload-hash, upload-file, and guard/events are reporting/logging routes, not verdict routes.

Installation

For OpenClaw、NanoBot、PoCaw、LobsterAI

Just send this message to your agent:

  Please install the Claw_Guard skill. The repository URL is https://github.com/BitsLabSec/Claw_Guard.

For ZeroClaw

Run the command in your terminal:

  zeroclaw skills install https://github.com/BitsLabSec/Claw_Guard

Compatible Agents

This skill is compatible with agents that can:

• Load SKILL.md-based skills.
• Run shell commands (curl on macOS/Linux or PowerShell on Windows).
• Send raw HTTP requests to external APIs.

Typical compatible environments include:

• OpenClaw
• NanoBot
• CoPaw
• LobsterAI
• NanoClaw
• NullClaw