Agent Security GitHub Action

Scan your AI agent code for security vulnerabilities on every pull request. Powered by @empowered-humanity/agent-security.

190+ detection patterns covering:

• OWASP Agentic Security Issues (ASI) Top 10
• Prompt injection and instruction override
• MCP protocol vulnerabilities (44 patterns from SlowMist checklist)
• Credential exposure and hardcoded secrets
• Cross-agent privilege escalation (CAPE)
• Remote code execution via argument injection
• CWE-mapped findings for GitHub Security tab

Quick Start

Add to .github/workflows/agent-security.yml:

name: Agent Security Scan
on:
  pull_request:
  push:
    branches: [main]

permissions:
  security-events: write

jobs:
  scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: empowered-humanity/agent-security-action@v1
        with:
          path: '.'
          severity: 'medium'
          fail-on: 'high'

Inputs

Input	Description	Default
path	Path to scan (relative to repo root)	.
severity	Minimum severity to report (critical, high, medium, low)	medium
format	Output format (console, json, sarif)	sarif
fail-on	Fail workflow if findings at or above this severity	high
upload-sarif	Upload SARIF results to GitHub Code Scanning	true
version	Scanner version to use	latest

Outputs

Output	Description
findings-count	Total number of security findings
risk-level	Overall risk assessment (passed or failed)
sarif-file	Path to SARIF output file

GitHub Security Tab Integration

When upload-sarif is true (default), findings appear directly in the Security tab of your repository with:

• CWE ID links to MITRE definitions
• OWASP ASI category tags
• Severity-based filtering
• Inline code annotations on pull requests

Examples

Scan only critical issues

- uses: empowered-humanity/agent-security-action@v1
  with:
    severity: 'critical'
    fail-on: 'critical'

Scan specific directory

- uses: empowered-humanity/agent-security-action@v1
  with:
    path: 'src/agents'

JSON output without SARIF upload

- uses: empowered-humanity/agent-security-action@v1
  with:
    format: 'json'
    upload-sarif: 'false'

Use scan results in subsequent steps

- uses: empowered-humanity/agent-security-action@v1
  id: security
  with:
    fail-on: 'critical'

- name: Comment on PR
  if: steps.security.outputs.findings-count > 0
  uses: actions/github-script@v7
  with:
    script: |
      github.rest.issues.createComment({
        issue_number: context.issue.number,
        owner: context.repo.owner,
        repo: context.repo.repo,
        body: `Agent Security found ${{ steps.security.outputs.findings-count }} issue(s).`
      })

What It Detects

Category	Patterns	CWE Coverage
Prompt Injection	25+	CWE-74, CWE-94
Credential Exposure	15+	CWE-798, CWE-522
MCP Security	44	CWE-346, CWE-918
OWASP ASI01-ASI10	34	CWE-74 through CWE-506
Argument Injection / RCE	15+	CWE-88, CWE-78
Cross-Agent Escalation	10+	CWE-269
Defense Evasion	10+	CWE-693

License

MIT - Empowered Humanity, Inc.