ClawSecCheck skill
An OpenClaw skill that audits the agent itself: 230 checks, change watching and an A to F grade.
- 59 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill reads the agent configuration and the skill list and changes nothing in them.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/clawseccheckDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill reads the agent configuration and the skill list and changes nothing in them.
This is third-party code. Review the repository files before installing.
What it does
ClawSecCheck audits an OpenClaw agent setup without changing its configuration. It works in chat: you ask to check your OpenClaw and the agent answers in plain words about what is dangerous and why. The scanner makes no network calls, needs no API key and sends nothing out. There are three modes. A full check runs 230 checks and 26 attack-chain detectors across five layers and issues an A to F grade only if every layer ran; otherwise it names the most urgent finding and the layers that did not run. Watch mode records a baseline and reports what changed: a new skill, a new MCP server, configuration drift. A pre-install gate inspects a new skill beforehand. Extras: a canary self-test for prompt injection, an evidence bundle for suspected compromise and a PDF report. Install with `openclaw skills install @gl0di/clawseccheck`.
Who it is for. OpenClaw agent owners who add third-party skills and MCP servers and want to know what is dangerous in that setup.
Good fit when
- A new skill or MCP server was added and you want to assess it before use.
- The agent holds keys and reads messages, and the setup has not been reviewed in a while.
- You want to see weekly what changed in the agent configuration.
- Compromise is suspected and you need to collect material for investigation.
Not a fit when
- The agent does not run on OpenClaw: the check is verified on version 2026.11.2 and on Linux.
- You need automatic remediation: the skill only reports.
- You expect a full audit with a guarantee of safety.
Example request
Audit my OpenClaw setup with clawseccheck and show the most urgent finding.Limitations
Verified on OpenClaw 2026.11.2 and Linux; older configuration shapes before 2026.8.1 are also read. It needs Python 3.10 or newer. The skill only reports what it finds and fixes nothing. An A to F grade is issued only after all five audit layers ran, and a default run usually reaches two of five and gives no grade. Two optional disk writes are made by the skill itself: a warning suppression file and a PDF report into the OpenClaw attachment directory. Static analysis does not catch every attack, so a clean report does not guarantee safety.
How to disable. Remove the skill with the OpenClaw uninstall command for `clawseccheck`, then delete the warning suppression file and the saved watch-mode baselines.
FAQ
How do I start a check?
Tell the agent to audit the setup with `clawseccheck`. No flags or commands are needed: findings arrive in chat, most urgent first.
Why is there no grade in my report?
An A to F grade needs all five layers. A default run reaches two and names the layers it skipped instead of printing a guessed grade.
What does watch mode do?
The first run records a baseline and says so. Later runs report only the difference: a new skill, a connected MCP server, a changed setting. `--verify-baseline` checks the baseline.
Does the skill send data out?
No, the scanner makes no network calls and sends no telemetry. No API key is needed.
Does it fix what it finds?
No, it only reports. The skill never changes the OpenClaw configuration.
Discussion
No questions or reviews yet.
Related
SkillsSecurity skills for OpenClaw, NanoClaw, Hermes and Picoclaw: signed advisories, guarded installs and drift detection
ClawScan runs several security scanners on agent skills and hands their results to a judge for a combined verdict.
Security scanner for agent skills and MCP servers: finds prompt injection, data exfiltration and supply chain risks before install
A CLI that scans your machine for skills and MCP configs, finds poisoned tool descriptions and runs a system prompt through 225 attacks
A Claude Code skill that finds vulnerabilities in one session: threat model, verified findings, CWE and CVSS report.
AgentShield audits the .claude folder for plaintext keys, broad permissions, dangerous hooks and risky MCP servers and grades the setup from A to F
README in short
A skill that audits OpenClaw agent security through chat and works offline. Modes: full check, change watching and a pre-install gate.