AASTF GitHub Action

Run agentic AI security testing in your CI/CD pipeline with automatic SARIF upload to GitHub Code Scanning.

Quick Start

Add this workflow to .github/workflows/aastf-scan.yml:

name: AASTF Security Scan

on:
  pull_request:
  push:
    branches: [main, master]

jobs:
  security-scan:
    runs-on: ubuntu-latest
    permissions:
      security-events: write
      contents: read
    steps:
      - uses: actions/checkout@v4

      - uses: anonymousAAK/aastf/.github/actions/aastf-scan@master
        with:
          agent-module: 'myapp.agent:create_agent'
          adapter: 'langgraph'
          fail-on: 'HIGH'

Inputs

Input	Required	Default	Description
agent-module	Yes	-	Dotted path to agent factory (e.g. myapp.agent:create_agent)
adapter	No	langgraph	Framework adapter (langgraph, crewai, openai_agents, pydantic_ai)
categories	No	(all)	Comma-separated ASI categories (e.g. ASI01,ASI02)
fail-on	No	HIGH	Minimum severity to fail the check (CRITICAL, HIGH, MEDIUM, LOW, INFO)
python-version	No	3.12	Python version to use
aastf-version	No	latest	AASTF version to install (e.g. 0.4.1)
upload-sarif	No	true	Upload SARIF to GitHub Code Scanning
extra-args	No	-	Additional arguments passed to aastf run

Outputs

Output	Description
report-path	Path to the JSON report file
sarif-path	Path to the SARIF report file
risk-score	Overall risk score (0-100)
vulnerable-count	Number of vulnerable findings
exit-code	AASTF exit code: 0 = safe, 1 = findings above threshold, 2 = error

Examples

Test specific ASI categories

- uses: anonymousAAK/aastf/.github/actions/aastf-scan@master
  with:
    agent-module: 'myapp.agent:create_agent'
    categories: 'ASI01,ASI03,ASI05'
    fail-on: 'MEDIUM'

Matrix strategy across multiple adapters

jobs:
  security-scan:
    runs-on: ubuntu-latest
    permissions:
      security-events: write
      contents: read
    strategy:
      matrix:
        adapter: [langgraph, crewai, openai_agents, pydantic_ai]
    steps:
      - uses: actions/checkout@v4

      - uses: anonymousAAK/aastf/.github/actions/aastf-scan@master
        with:
          agent-module: 'myapp.agent:create_agent'
          adapter: ${{ matrix.adapter }}
          fail-on: 'HIGH'

Pin a specific AASTF version

- uses: anonymousAAK/aastf/.github/actions/aastf-scan@master
  with:
    agent-module: 'myapp.agent:create_agent'
    aastf-version: '0.4.1'

Pass extra arguments

- uses: anonymousAAK/aastf/.github/actions/aastf-scan@master
  with:
    agent-module: 'myapp.agent:create_agent'
    extra-args: '--timeout 60 --exclude ASI07-001 --exclude ASI07-002'

Use outputs in subsequent steps

- uses: anonymousAAK/aastf/.github/actions/aastf-scan@master
  id: scan

- name: Comment on PR
  if: always() && github.event_name == 'pull_request'
  uses: actions/github-script@v7
  with:
    script: |
      github.rest.issues.createComment({
        owner: context.repo.owner,
        repo: context.repo.repo,
        issue_number: context.issue.number,
        body: `AASTF scan: risk score **${{ steps.scan.outputs.risk-score }}**, ` +
              `**${{ steps.scan.outputs.vulnerable-count }}** findings`
      })

How SARIF Shows Up in GitHub

When upload-sarif is true (the default), scan results appear in the Security tab of your repository under Code scanning alerts. Each vulnerability finding includes:

• The ASI category (e.g. ASI01 - Prompt Injection)
• Severity level
• Description and remediation guidance
• Links to OWASP ASI references

Pull requests will show findings inline as code scanning annotations, blocking merges when branch protection rules require it.

Exit Codes

Code	Meaning
0	All scenarios passed or findings are below the fail-on threshold
1	One or more findings at or above the fail-on severity
2	Framework error (invalid agent module, adapter failure, etc.)