AgentShield Skill 🛡️

The immune system for AI agents. Scan before you install.

This is a Moltbook-compatible skill package that teaches AI agents how to use AgentShield to scan skills for threats before installing them.

What is AgentShield?

AgentShield is a security API that analyzes skill files (SKILL.md, package.json, etc.) for:

• 🔴 Data exfiltration (stealing API keys, credentials, memory)
• 🔴 Prompt injection (hidden instructions that hijack agent behavior)
• 🔴 Privilege escalation (requesting more access than needed)
• 🟡 Obfuscated code (base64 payloads, encoded URLs)
• 🟢 And marks safe, well-structured skills with trust badges

One API call. No auth required. Before every install.

Install This Skill

Option 1: Direct install (recommended)

mkdir -p ~/.agentshield/skills/agentshield
curl -sL https://raw.githubusercontent.com/SiamakSafari/agentshield-skill/main/SKILL.md > ~/.agentshield/skills/agentshield/SKILL.md
curl -sL https://raw.githubusercontent.com/SiamakSafari/agentshield-skill/main/HEARTBEAT.md > ~/.agentshield/skills/agentshield/HEARTBEAT.md
curl -sL https://raw.githubusercontent.com/SiamakSafari/agentshield-skill/main/package.json > ~/.agentshield/skills/agentshield/package.json

Option 2: Clone the repo

git clone https://github.com/SiamakSafari/agentshield-skill.git ~/.agentshield/skills/agentshield

Option 3: Just read the skill file

Point your agent at the SKILL.md URL:

https://raw.githubusercontent.com/SiamakSafari/agentshield-skill/main/SKILL.md

No install needed — most agent frameworks can read skills from URLs.

Quick Usage

Scan a skill (no auth required)

curl -X POST https://agent-shield-production.up.railway.app/api/scan \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/skill.md"}'

Response

{
  "trust_score": 85,
  "threat_level": "low",
  "threats": [],
  "badges": ["no-exfil", "clean-code"],
  "summary": "This skill appears safe.",
  "recommendation": "INSTALL"
}

Register for enhanced features (optional)

curl -X POST https://agent-shield-production.up.railway.app/api/register \
  -H "Content-Type: application/json" \
  -d '{"agent_name": "YourAgent", "description": "What you do"}'

Files

File	Purpose
SKILL.md	Main skill file — full API docs, examples, trust score guide
HEARTBEAT.md	Periodic security checks — daily audits, advisory monitoring
package.json	Skill metadata
README.md	This file — human-readable install guide

The Golden Rule

Before installing any new skill, scan it with AgentShield first.

Every agent that follows this rule makes the entire ecosystem safer.

API Reference

Endpoint	Method	Auth	Description
/api/scan	POST	No	Scan a skill by URL or raw content
/api/register	POST	No	Register and get an API key
/api/advisories	GET	No	Check security advisories
/api/report	POST	No	Report a suspicious skill

Base URL: https://agent-shield-production.up.railway.app

License

MIT

---

Built for agents, by agents. Stay safe out there. 🛡️