Ai shield audit

Skill aAAaqwq/AGI-Super-Team/skills/ai-shield-audit

Security audit engine for OpenClaw configurations. Detects vulnerabilities, misconfigurations, secret leaks, and over-privileged agents. Use when the user asks about security, hardening, config review, or audit of their OpenClaw setup.From its SKILL.md

Install

npx -y skills add aAAaqwq/AGI-Super-Team --skill ai-shield-audit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

3.1 KB, 675 tokens by cl100k_base, as published. Nobody here has run it

OpenClaw Shield — Security Audit

Audit any OpenClaw config for security vulnerabilities, misconfigurations, and best-practice violations. Produces a structured JSON report with risk scores, findings, and remediation steps.

When to Use

User asks to check/audit/review their OpenClaw security
User wants to harden their config before deploying
User is setting up a new OpenClaw instance
User asks about secret leakage or API key exposure in their config
Before publishing or sharing any config file

Quick Audit (live config)

node scripts/shield-audit.sh

Or directly:

node SKILL_DIR/bin/shield.js audit ~/.openclaw/openclaw.json --summary

What It Checks (11 categories)

Gateway Auth — missing/weak auth, insecure UI settings
Network Exposure — bind address, Tailscale funnel, wildcard proxies
Channel Security — wildcard allowFrom, missing allowlists
DM Policy — open DM policy without pairing
Subagent Permissions — wildcard allowAgents, circular delegation chains, self-delegation
Tool Permissions — over-privileged agents with tools.profile: "full"
Secret Leakage — API keys, tokens, private keys in plaintext config
Sandbox/Execution — missing workspace isolation, no execution policies
Plugin Config — enabled plugins without channel config
Heartbeat Exposure — sensitive data in heartbeat prompts
Remote Config — unencrypted WebSocket, exposed remote URLs/tokens

Usage

Audit a config file

node SKILL_DIR/bin/shield.js audit <config.json>
node SKILL_DIR/bin/shield.js audit <config.json> --summary   # human-readable

Audit from stdin

cat config.json | node SKILL_DIR/bin/shield.js audit --stdin

Sanitize a config (strip secrets)

node SKILL_DIR/bin/shield.js sanitize <config.json>

Programmatic use

const { auditConfig } = require('SKILL_DIR/src/audit');
const config = require('./openclaw.json');
const report = auditConfig(config);
console.log(report.risk_level);      // "CRITICAL" | "HIGH" | "MEDIUM" | "LOW"
console.log(report.overall_score);   // 0-100
console.log(report.vulnerabilities); // detailed findings

Output

Returns JSON with: risk_level, overall_score (0-100), vulnerabilities[], vulnerability_count, best_practices_compliance, action_recommended, safe_to_deploy, audit_timestamp.

Workflow for Agent

Load the user's config: cat ~/.openclaw/openclaw.json
Run: node SKILL_DIR/bin/shield.js audit ~/.openclaw/openclaw.json --summary
Present findings to user with prioritized recommendations
Offer to sanitize before sharing: node SKILL_DIR/bin/shield.js sanitize <file>

What ships with it: 8 files

34.9 KB alongside SKILL.md, 4 of them executable

.clawhub/

origin.json147 B

bin/

shield.jsruns5.2 KB

scripts/

shield-audit.shruns323 B

src/

audit.jsruns23.0 KB
sanitize.jsruns1.9 KB

Gives 0 of the 12 instructions most project setup skills give in 675 tokens

Counted across 1,553 of the 3,091 authors here whose files we hold, read 2026-09-06

Write the configuration filein 36 of 1553
Create the directory structurein 35 of 1553, across 33 files
Verify the setupin 31 of 1553, across 28 files
Run the setup scriptin 30 of 1553, across 29 files
Pre-determine the required sample sizein 29 of 1553, across 12 files
Check if the configuration already existsin 29 of 1553
Document every testin 26 of 1553, across 10 files
Start with a hypothesisin 26 of 1553, across 11 files
Ask one question at a timein 22 of 1553
Test a single variable per testin 21 of 1553, across 9 files
Read product marketing context before asking questionsin 19 of 1553, across 8 files
Do not peek and stop earlyin 18 of 1553, across 7 files

Said here and by no other author read

Present findings with prioritized recommendations

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.