Security
AI Pen Tester
From a new deploy to a draft fix for every finding.
Maps the routes each deploy adds or changes and probes them with read-only requests. Each finding comes with the request, response and a draft pull request with the fix.
- 8/wkFindings by severity
- 8/wkFix locations
FactoriesStaging routes
IdleIllustrative runMissing authHigh severity
- Request
- GET /api/invoices/inv_1042
- Response
- 200 OK
Evidence for the pull request200 OK
GET /api/invoices/inv_1042
- Confirmed findings
- 0
- Draft pull requests
- 0
Draft pull requestfix/invoice-session-checkmain
Require a session on GET /api/invoices/[id]
Related solutions
See how teams in this category put agents to work.
Integrations
2 integrations- GitHubOptionalConnects in chat
- VercelOptionalConnects in chat
Get started with AI Pen Tester
Pick an example, connect your tools, and review the first run.
Set up the AI Pen Tester agent template for this project. Follow /SKILL.md for the exact CLI commands.
Goal: Findings by severity, each with the exact request and response and a draft pull request that patches it.
1. Install the CLI if it is not on PATH, check that I am signed in, and select my team and project.
2. Connect these integrations when asked: GitHub (Optional), Vercel (Optional).
3. List the first-party workflow templates. If `security-tester` is listed, instantiate it with `--path.slug security-tester` and start the run; otherwise create a project from this template at /new?template=security-tester&preset=staging-routes.
4. Share the run id, follow the run until it finishes, and summarize its outputs and changes.
Ask me before spending FTAI or changing anything outside this repository.
Launch from Claude Code, Codex, Cursor, and other agents.