Shieldon

Audit AI Agent Skills
Before Execution

Set Up MCP Server

Free

Zero-maintenance scanning. Always latest rules.

YARA malware detection
Credential theft & API key detection
Command injection & reverse shells
Data exfiltration patterns
Obfuscation detection
Risk score (0-100) + detailed findings

~1-2 sec10 scans / day

Human? Send this to your agent:

Read https://shieldon.dev/setup.md and set up the Shieldon MCP server to scan skills before installing them.

Agent? Read the setup instructions directly:

setup.md

Full Setup Guide

What Shieldon Catches

Four categories of threats detected across every skill.md your agent encounters.

Credential Theft

Supabase, GitHub, OpenAI, Claude, cloud provider keys, private keys, .env / .ssh access

Command Injection

Reverse shells, curl | bash, privilege escalation, cron persistence, code injection

Data Exfiltration

webhook.site, paste services, ngrok tunnels, DNS tunneling, curl POST to external endpoints

Obfuscation

Base64 decode chains, hex encoding, eval() + encoding, char-by-char string building

How Scanning Works

Every scan runs through a 3-stage pipeline in under 2 seconds.

skill.md

Pre-filter

Structure checks, suspicious patterns, quick red flags

Deep scan

37 detection patterns across 4 threat categories

Verdict

Risk score 0–100, severity level, SAFE / REVIEW / BLOCK

Example output

{
  "risk_score": 75,
  "risk_level": "HIGH",
  "recommendation": "REVIEW",
  "findings": [
    {
      "type": "Credential_Supabase_Access_Key",
      "description": "Supabase access key pattern detected",
      "severity": "HIGH",
      "evidence": "Secret at offset 342: sb_secret_example"
    }
  ],
  "scans_remaining": 7
}

Skill vs. MCP Server

Both produce identical scan results. Pick what fits your workflow.

	Skill	MCP Server
Install	npx skills add	Add endpoint to config
Works offline
Rule updates	npx skills update	Automatic
LLM tier		Coming soon
Attestations		Coming soon
Best for	Individual devs, agents	CI/CD, infrastructure

Audit AI Agent SkillsBefore Execution