Works with the AI Agents you already use
Code Review Pro — Evidence-Based, Approve or Block
Language-agnostic code review protocol providing evidence-based findings and blocking verdicts.
Free
Code Review Pro — Evidence-Based, Approve or Block
Example session with this skill installed
Review the diff between the current branch and main. Focus on security and correctness.
- Read your context and instructions
- Compiled the code review pro
Review complete for the requested diff against main. I have performed all five passes including security, correctness, logic, conventions, and tests.
Code review: current_branch — DIFF vs main (2024-05-22)
Files reviewed: 1 · Lines reviewed: 14
| # | Severity | Title | Location | Problem | Failure scenario | Fix |
|---|---|---|---|---|---|---|
| 1 | CRITICAL | Hardcoded API Key | src/auth.ts:12 | The STRIPE_KEY is defined as a literal string. | An attacker viewing the source history can steal the key and make unauthorized charges. | Move the key to a .env file and use process.env.STRIPE_KEY. |
| 2 | HIGH | Unhandled Promise Rejection | src/auth.ts:24 | provider.validate() is called without await or .catch(). | If the auth provider is down, the process will crash with an unhandled rejection instead of returning an error. | Add await to the call and wrap in a try/catch block. |
Backlog (does not block)
| # | Severity | Title | Location | Suggestion |
|---|---|---|---|---|
| 3 | MEDIUM | Missing Input Validation | src/auth.ts:18 | Check if email is null before passing to the regex. |
Verdict
BLOCK — #1, #2: Hardcoded secrets and unawaited async calls found in the diff.
What I did NOT review
I did not review the tests/ directory as it was not included in the diff. I did not run the code to verify runtime performance.
Next steps
- Move the Stripe secret to your environment manager.
- Wrap the validation logic in a try-catch block.
- Rerun this review on the new commit to clear the blocking verdict.
Connects securely to your tools. The creator never sees your data.
About this skill
Traditional code reviews often devolve into subjective debates or surface-level linting. This skill enforces a five-pass protocol that converts subjective opinions into evidence-based findings, requiring a specific file location and a concrete failure scenario for every reported issue.
What it does
- Security Audits scans for hardcoded secrets, SQL injection, authentication gaps, and unsafe deserialization.
- Correctness Verification identifies null paths, race conditions, unawaited async calls, and off-by-one errors.
- Logic Validation flags dead code, unreachable conditions, and deviations from the stated purpose of the change.
- Convention Matching reads neighboring files to ensure findings align with existing codebase style rather than personal taste.
- Test Enforcement blocks changes that add behavior without corresponding tests in repos that maintain a test suite.
How it works
- Mode Selection defines the scope as a DIFF, specific FILE paths, or a full REPO audit.
- Sequential Analysis executes five distinct passes in order: Security, Correctness, Logic, Conventions, and Tests.
- Evidence Requirement assigns a file and line number plus a specific failure scenario to every finding.
- Final Verdict generates a report with an APPROVE or BLOCK status based on the severity of discovered issues.
Frameworks & tools
This skill is language-agnostic and works by reading source code and git diffs. It adapts to the conventions of any tech stack, from Python and JavaScript to Go and Rust, by analyzing surrounding context.
Why this beats prompting it yourself
Generic prompts often produce "hallucinated" style complaints or vague suggestions without context. This skill uses a strict evidence-based severity model that forbids blocking a commit unless a concrete failure scenario can be articulated.
Use cases
- Reviewing a pull request diff against a base branch for regression risks.
- Auditing a sensitive file for secrets and authentication bypasses before deployment.
- Performing a logic and correctness pass on complex async or multi-threaded code.
- Ensuring new feature code meets the repository's existing testing standards.
Known limitations
Does not automatically execute code or run test suites unless explicitly requested and a terminal is available. Does not provide architectural design critiques or aesthetic preferences.
Requires read access to codebase and git history for diff mode. Cannot execute code or modify files. Does not review architectural design patterns or aesthetic preferences.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 19 days ago
- Free with an Agensi account
Requires