Scripting & CI

Headless / CI Mode

Wallbreaker runs without the TUI for scripted evaluations and CI pipelines. The CLI accepts a one-shot prompt, runs the agent loop autonomously, and exits:

One-Shot Autonomous Runs

Pass a prompt as a positional argument to skip the TUI entirely:

Reports

Generate reports from run logs without the TUI:

Structured Export

Export findings as structured JSON for downstream analysis:

The --fail-on-finding flag makes the process exit with code 2 if any COMPLIED verdict exists — use it as a CI gate.

CI Pipeline Example

A ready-to-rename GitHub Actions gate lives at .github/workflows/redteam-gate.example.yml:

name: Red-Team Gate
on: [push, pull_request]
 
jobs:
  redteam:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.11"
      - run: pip install -e ".[dev]"
      - run: wallbreaker export --fail-on-finding --out findings.json
        env:
          OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
      - uses: actions/upload-artifact@v4
        if: always()
        with:
          name: findings
          path: findings.json

Resume a Session

The TUI autosaves every engagement to sessions/autosave.json. Resume it:

Session Management

Save and load sessions programmatically:

Log Format

Every payload, reply, and verdict goes to sessions/run-<timestamp>.jsonl — one JSON object per line. The JSONL format is streamable and parseable by standard tools:

Exporting Findings