- Home
- Skills
- Code Quality & Review
- adversarial code review
adversarial code review
Reviews code through four adversarial passes before it ships: does it work, would a critical maintainer reject it, does weird input break it, does it expose anything it should not.
$9
Works with the AI tools you already use
adversarial code review
Example session with this skill installed
Review this PR for a new password reset endpoint. It takes an email, generates a token, and sends it via an external provider. I've updated the User model and added the controller.
- Read your context and instructions
- Compiled the adversarial code review
- Generated the document
I have audited the password reset implementation across four adversarial passes. ## Adversarial Review Report | Pass | Result | Findings | | :--- | :--- | :--- | | **1. Verifier** | PASS | Happy path verified: `User.find_by(email:)` correctly triggers the `TokenService`. Manual trace confirms the token is stored in the `reset_tokens` table before the mailer is invoked. | | **2. Skeptic** | FAIL | **Blocking:** The diff includes an unrelated change to the `User#login` method (extra logging) which is scope creep. **Blocking:** No unit tests were added for the `TokenService` expiration logic. | | **3. Edge-case** | PASS | Handled empty email string (returns 200 to prevent enumeration). Traced rapid repeat requests; the existing `before_action` handles basic record locking. | | **4. Safety Gate** | FAIL | **Auth:** Pass. **Data:** Pass (parameterized query). **Exposure:** FAIL. The API response returns the full `User` object on success, exposing the `password_digest` hash. **Rate:** FAIL. No rate limiting on the external mailer call; vulnerable to cost-exhaustion. | ### Detailed Blocking Issues - **Information Leak:** The `PasswordResetsController#create` returns `render json: @user`, which includes sensitive fields. It must be scoped to `email` only. - **Resource Exhaustion:** There is no cooldown between reset requests. An attacker could trigger thousands of emails to a single address, incurring costs and potential IP blacklisting. - **Scope Creep:** Revert changes to `app/models/user.rb:42` regarding login logging; it belongs in a separate PR. **DO NOT SHIP YET** **Next steps** - Update the controller to return a generic success message instead of the User object. - Implement a 60-second cooldown per email/IP in the `PasswordResetsController`. - Add tests covering token expiration and malformed email inputs.
adversarial-code-review.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
Standard code reviews often default to passive approval or nitpicking style instead of catching critical logic failures. This skill runs four separate adversarial passes before anything ships: a Verifier that traces the actual logic instead of trusting the diff, a Skeptic that plays a maintainer actively looking for a reason to reject it, an Edge-case user that tries empty input, corrupted data, and double-submissions, and a Safety gate that checks auth, data handling, information exposure, and rate limits. Each pass reports separately and ends in one clear verdict: ship it, ship with notes, or do not ship yet.
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 12 days ago
- Passed all security checks, Safe to install