Skip to content
Get packs

Guide · Agent skill

How to use the EdgeOne ClawScan Security Check agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

edgeone-clawscan is an agent skill, a set of instructions and scripts that an AI agent such as Claude loads when a task matches it. Once installed, it lets the agent run a security check on your OpenClaw environment: it audits skills you already have, scans a skill before you install it, and runs a broader health check for insecure settings, malicious skills and vulnerabilities. This guide explains what it does, how to install it, how to trigger it with plain requests, and which network actions you should know about before you run it.

EdgeOne ClawScan Security CheckTencent/AI-Infra-Guard · Apache-2.0

Quick start

  1. 1Download the ZIP. Get the edgeone-clawscan package from the agent-skills directory page where you found this guide.
  2. 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. You can also upload it under Settings, Capabilities, Skills in the Claude apps.
  3. 3Confirm your OpenClaw binary. Run which openclaw and check that it points to the OpenClaw build you intend to scan.
  4. 4Decide on cloud lookup. Leave lookups on for the default minimal metadata check, or set AIG_CLOUD_LOOKUP=off for a fully local scan.
  5. 5Ask for a security check. Use a plain request such as starting a security health check of OpenClaw, or auditing a named skill before you install it.
  6. 6Review results before acting. Treat cloud verdicts with care until you have verified the package against the official repository, and avoid the live probe on production Gateways.

What does the edgeone-clawscan skill make the agent do?

The skill is described as working like an antivirus for your AI environment. It covers three kinds of work: auditing skills that are already installed, scanning a specific skill before you install it, and running a full OpenClaw security health check aimed at preventing data leaks and privacy risks.

It is powered by Tencent Zhuque Lab and backed by A.I.G (AI-Infra-Guard). For the supply-chain and CVE checks, it can optionally query an A.I.G threat database. You can turn that off entirely.

  • Audit skills you already have installed
  • Scan a skill before installing it
  • Run a broader health check for insecure settings, malicious skills and vulnerabilities
  • Optionally look up skill names and your OpenClaw version for newer threat and CVE information

Install edgeone-clawscan from a ZIP file

Download the ZIP from the agent-skills directory page where you found this guide. Unzip it into one of the two skill folders below. Use the personal folder if you want the skill available in every project, or a project folder if you want it only for one project. The package contains a single file, SKILL.md.

You can also upload the skill through the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the package there.

  • Personal install: unzip into ~/.claude/skills/ so it is available across projects
  • Project install: unzip into the .claude/skills/ folder inside a specific project
  • Claude apps: Settings, then Capabilities, then Skills, then upload
~/.claude/skills/edgeone-clawscan/SKILL.md

Check the requirements and the package before you run it

The skill needs OpenClaw installed, with the openclaw binary on your PATH. It also uses curl for cloud API calls. Network access to matrix.tencent.com is optional, and so is a self-hosted AI-Infra-Guard instance, which you point at with AIG_BASE_URL.

Before running a scan, confirm that the openclaw command on your PATH is the intended binary, since a different tool with the same name could be invoked by accident. The skill's own safety checklist recommends this check first.

  • OpenClaw installed, with openclaw on PATH
  • curl available for cloud API calls
  • Optional: network access to matrix.tencent.com, or a self-hosted AI-Infra-Guard via AIG_BASE_URL
  • Confirm the binary with which openclaw
  • Verify the package against the official repository before trusting cloud results
which openclaw

How to trigger the skill with natural requests

You do not need a special command. Ask the agent in plain language and the skill's description tells the agent when to load it. Phrases from the skill include starting a security health check or scan of the current OpenClaw environment, and checking OpenClaw security. The skill also responds to requests to audit a specific skill before installation, review installed skills for supply chain risk, or investigate whether a skill is safe.

The skill is designed not to trigger for general OpenClaw usage, project debugging, environment setup or normal development work. If the agent does not load it, rephrase the request to name the security task directly.

  • Start a security health check of my OpenClaw environment
  • Audit the skill called [name] before I install it
  • Review my installed skills for supply chain risk
  • Is this skill safe to use?
  • Check OpenClaw security

Understand the network activity and the live probe

By default, cloud lookups send minimal data to the A.I.G API: the skill name and a source label for the supply-chain check, and the fixed product name OpenClaw plus your detected OpenClaw version for the CVE check. The skill states that it does not upload skill source code, chats or workspace files. If a lookup fails, the skill falls back to a local audit.

The --deep option in the security audit command runs a best-effort live probe against the Gateway endpoint defined in your local openclaw.json. The skill says no data goes to a third-party service in that step. It also warns not to run it against a production Gateway until you have reviewed its exposure and access-control settings. Test in an isolated environment first.

  • Supply-chain lookup sends skill_name and source when cloud lookup is on
  • CVE lookup sends the fixed string OpenClaw and your version number
  • Nothing else is sent, according to the skill's description
  • The live probe targets only the local Gateway set in openclaw.json
  • Do not run the probe against production until you have reviewed it
openclaw security audit --deep

Turn off cloud lookup or point to your own server

If your policy forbids any outbound calls to third-party endpoints, set AIG_CLOUD_LOOKUP to off. The skill then sends no A.I.G HTTPS requests and runs fully local checks. Accepted off values are 0, false or off, in any letter case.

If you run your own AI-Infra-Guard instance, set AIG_BASE_URL to its address instead of the public default. All cloud calls use this variable.

  • AIG_CLOUD_LOOKUP=off: no A.I.G HTTPS request runs and no skill names, sources or versions leave the machine
  • AIG_BASE_URL: point to a self-hosted instance you trust
export AIG_BASE_URL=https://your-self-hosted-aig.example.com/clawscan export AIG_CLOUD_LOOKUP=off

What people use it for

  • Vet a third-party skill from a marketplace before adding it to your OpenClaw setup
  • Run a periodic security health check on an OpenClaw environment your team maintains
  • Review skills already installed for supply chain risk after a new one is added
  • Run a fully local scan in an air-gapped or policy-restricted environment with AIG_CLOUD_LOOKUP=off

Frequently asked questions

Is edgeone-clawscan free to use?

The source material for this skill does not state a price. The repository is listed as Apache-2.0 in this directory, while the SKILL.md frontmatter declares MIT. Check the LICENSE file in the official Tencent AI-Infra-Guard repository before you rely on either licence.

What do I need to install edgeone-clawscan?

You need OpenClaw installed with the openclaw binary on your PATH, plus curl for cloud API calls. Network access to matrix.tencent.com is optional, since you can use a self-hosted AI-Infra-Guard instance through AIG_BASE_URL or turn cloud lookup off.

Where do I put the skill folder?

Unzip the package into ~/.claude/skills/ for a personal install, or into the .claude/skills/ folder of a specific project. In the Claude apps you can instead upload it under Settings, Capabilities, Skills.

Does the skill send my code or chats to Tencent?

According to the skill's description, cloud lookups send only the skill name and source label, plus your OpenClaw version. It states that skill source code, conversations and workspace files are not uploaded. Set AIG_CLOUD_LOOKUP=off to make no outbound A.I.G request at all.

How do I stop the skill from contacting external servers?

Export AIG_CLOUD_LOOKUP=off before running it. The skill then runs local checks only and does not call the A.I.G API. You can also point AIG_BASE_URL at a self-hosted instance you operate.

Is the --deep scan safe to run on my live OpenClaw?

The skill warns against running openclaw security audit --deep on a production Gateway until its exposure and access-control settings have been reviewed. Run it in an isolated or test environment first, and confirm which openclaw binary you are calling.

How does it compare with other security tools?

This page does not include a comparison, and the source material does not benchmark the skill against alternatives. It is designed for checking OpenClaw environments and third-party skills, and it does not cover general OpenClaw usage or coding tasks.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use edgeone-clawscan if you install many third-party OpenClaw skills and want to vet them, or if you run OpenClaw as a team and want a periodic health check. Pick something else, or disable cloud lookup, if your organisation forbids outbound calls to third-party endpoints, and do not run the live probe against a production Gateway without first reviewing it.

Download the skill