ClawHub

OpenClaw Tool Audit

v0.1.2

Audit OpenClaw agent tool exposure versus observed use. Use when reviewing allowed tools, spotting broad or unused tool allowances, or checking whether agent...

⭐ 0· 157·0 current·0 all-time
byPaul Frederiksen@pfrederiksen

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for pfrederiksen/openclaw-tool-audit.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "OpenClaw Tool Audit" (pfrederiksen/openclaw-tool-audit) from ClawHub.
Skill page: https://clawhub.ai/pfrederiksen/openclaw-tool-audit
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install openclaw-tool-audit

ClawHub CLI

Package manager switcher

npx clawhub@latest install openclaw-tool-audit
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
ℹ
Purpose & Capability
The skill's name and description match the runtime instructions: it expects a local openclaw-tool-audit binary to inspect OpenClaw configs and sessions and produce JSON/markdown reports. Slight metadata inconsistency: the top registry summary lists no required binaries, while skill.json/prerequisites declare the openclaw-tool-audit binary — this is explainable (instruction-only skill that relies on a local binary) but should be clarified.
✓
Instruction Scope
SKILL.md confines actions to running the local audit tool against local OpenClaw config and session data (which is necessary for the stated audit purpose). The document explicitly warns to verify the local binary and to avoid elevated privileges; it does not instruct the agent to exfiltrate data or contact external endpoints.
✓
Install Mechanism
There is no install spec and no code files — the skill is instruction-only. SKILL.md explicitly advises against remote installs and prefers preinstalled/trusted local binaries, which reduces install risk.
✓
Credentials
The skill requests no environment variables, credentials, or config-path access beyond the local OpenClaw config and session files that are required for the audit task. This access is proportional to the stated purpose.
✓
Persistence & Privilege
The skill is not always-enabled and does not request persistent privileges or modify other skills or system-wide settings. Autonomous invocation is allowed by the platform default, but there are no additional privilege escalations requested by the skill itself.
Assessment
This skill is internally consistent and low-risk if you follow its own safety guidance: only run it with a local openclaw-tool-audit binary you trust (inspect/build the source if unsure), avoid running as root, run with --json first to inspect output, and do not paste audit output containing session tokens or secrets to external services. Clarify the small metadata mismatch (registry summary vs skill.json) before relying on automated tooling that checks prerequisites. If you need higher assurance, review the referenced GitHub repo and build the binary locally in an isolated environment before use.

Like a lobster shell, security has layers — review code before you run it.

latestvk978xyynaekce44n6naqhchf91858ykk
157downloads
0stars
3versions
Updated 6d ago
v0.1.2
MIT-0
Paul Frederiksen@pfrederiksen
latest
Download

OpenClaw Tool Audit

Use this skill to audit local OpenClaw agent tool configuration against observed tool usage.

Repository

Primary source repo:

Prerequisites

Required:

  • a trusted local installation of openclaw-tool-audit
  • access to the local OpenClaw config and session data the tool reads

Before running:

  • verify the local binary or source install is one you trust
  • inspect local source if you did not build or install it yourself
  • avoid elevated/root execution unless you actually need it
  • confirm local session/config files do not expose secrets you are unwilling to inspect

When to use

Use this when the user asks to:

  • audit which tools agents are allowed to use
  • compare allowed tools vs tools actually used
  • spot overly broad tool access
  • review whether agent tool configs could be tightened
  • generate a markdown or JSON tool exposure report

Safe source guidance

Prefer one of these:

  • a previously installed trusted local binary on PATH
  • a trusted local source checkout you have already inspected
  • a pinned internal/local install workflow you control

Do not instruct users to install directly from a remote GitHub URL inside this skill.

Recommended commands

Default markdown summary:

openclaw-tool-audit --markdown --top-tools 15

JSON output:

openclaw-tool-audit --json

Broadest-first review:

openclaw-tool-audit --markdown --broadest-first

Unused-only review:

openclaw-tool-audit --markdown --unused-only

If the binary is not on PATH, use the trusted local path you already manage.

Parser / transcript notes

Real-world OpenClaw installs may have nested tool config shapes and transcript variants that require recent upstream fixes. If the tool crashes or reports zero observed invocations unexpectedly:

  • verify you are using a trusted current local build
  • validate with --json
  • check whether observed agent names and tool tokens look sane before making policy decisions

Recommended interpretation

Use the report to answer:

  • which agents have very broad allowlists
  • which tools are actually used most often
  • which agents have a high unused-allowance ratio
  • whether any observed tool tokens look like parser mistakes or unmatched config/runtime shapes

Packaging / safety

Keep this skill minimal and transparent:

  • plain text only
  • no binaries
  • no obfuscation
  • no remote install commands in SKILL.md
  • prefer already-installed local/auditable tooling

Comments

Loading comments...