Ask how the plan could fail.
Give AI reviewers an architecture, contract, or technical plan and ask them to look for failure modes. Treat the findings as questions to test, not a completed audit.
Architecture RFC & Design Review
Catch systemic failure modes before the first commit is written.
Include the design, dependencies, trust boundaries, expected load, and rollout plan. Ask reviewers to name a failure mechanism, likely impact, and a way to test the concern.
Reviewers explore the happy path, failure cases, and possible responses. The final conclusion brings their findings together so the responsible team can decide what to investigate, test, or change.
How It Works
Upload the relevant document and choose Red Team as the review format. Models take attack, defence, and assessment roles according to the configured panel.
Read the model responses and final conclusion, then verify each material finding. This process does not replace penetration testing, legal review, or a security professional.
Key Features
- bug_reportVulnerability Discovery: Identifies unhandled exceptions, race conditions, and logical flaws.
- shieldSecurity Focused: Prompts models to act as malicious actors searching for exploits rather than helpful assistants.
- speedFocused follow-up: Turn useful findings into tests and questions for the responsible reviewer.