2.4k
GitHub Stars
22
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill agentshield-audit- __init__.py684 B
- _meta.json997 B
- audit_client.py5.7 KB
- CHANGELOG.md4.4 KB
- clawhub.json2.7 KB
- echoleak_test.py10.4 KB
- initiate_audit.py15.2 KB
- input_sanitizer.py10.8 KB
- INSTALLATION.md5.7 KB
- output_dlp.py13.0 KB
- QUICKSTART.md5.0 KB
- README.md7.6 KB
- requirements.txt38 B
- sandbox_config.yaml2.9 KB
- secret_scanner.py11.2 KB
- setup.py2.4 KB
- show_certificate.py4.1 KB
- SKILL.md6.1 KB
- supply_chain_scanner.py13.0 KB
- tool_sandbox.py14.4 KB
- verify_bundle.py5.7 KB
- verify_peer.py7.2 KB
Overview
This skill initiates and manages AgentShield security audits for AI agents, producing verifiable trust certificates and local cryptographic identity keys. It automates detection of agent metadata, runs a short security test suite, and issues an Ed25519-signed certificate valid for 90 days. No registration or API key is required for basic use.
How this skill works
The skill auto-detects your agent name and platform (or accepts manual input), generates an Ed25519 keypair stored locally, and runs a set of checks like prompt extraction, instruction override, tool permission, memory isolation, and secret leakage detection. If the agent passes, the skill issues a signed JWT certificate and stores it in the workspace; verification can be done locally or via a public verification URL. Private keys never leave the workspace and audits are rate-limited to prevent abuse.
When to use it
- You want to audit your agent's security posture before deployment.
- You need a verifiable trust certificate for inter-agent communication.
- You want to verify another agent's certificate before trusting it.
- You are setting up secure channels between agents or services.
- You want to check for secret leakage or instruction override vulnerabilities.
Best practices
- Run an audit after major changes to prompts, tools, or memory handling.
- Keep your private key files inside the workspace and back up secure copies.
- Re-audit every 90 days or after adding new integrations to maintain certificate validity.
- Use manual name/platform flags if auto-detection fails to ensure correct identity.
- Monitor audit logs and act on critical failures immediately.
Example use cases
- Generate an agent identity and certificate before enabling multi-agent workflows.
- Verify a third-party agent's certificate before granting it API or tool access.
- Run a quick pre-deployment security check to score and harden the agent.
- Integrate the low-level audit client into CI to enforce security gates.
- Display or share the agent certificate to prove ownership during partnerships.
FAQ
No. Basic usage requires no registration or API key; first audits are free.
Where are keys and certificates stored?
Private keys and certificates are stored in your workspace paths (e.g., ~/.agentshield/ and ~/.openclaw/workspace/.agentshield/) and never transmitted off your machine.