code-review

Reviews code implementation against Product-Spec and design specs with evidence-backed reports.

Updated Jul 3, 2026
One-click install
npx skills add https://github.com/LYGOLANG/fufan-cc-flow --skill code-review-lygolang
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/LYGOLANG/fufan-cc-flow/tree/main/.agents/skills/code-review
Command: npx skills add https://github.com/LYGOLANG/fufan-cc-flow --skill code-review-lygolang

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Verifying that code actually implements what the product spec promised is tedious and error-prone. This Skill performs a rigorous, evidence-based audit of your codebase against Product-Spec.md and design files, so no missing feature, fake test, or security flaw slips through unverified claims. ## Core Features & Use Cases - Two-Stage Review: Stage 1 checks functional completeness, UI guidance authenticity, and design consistency; Stage 2 covers code quality, test authenticity, security scanning, spec drift, and visual comparison. - Evidence-First Reporting: Every conclusion cites file line numbers, compile output, or API responses; issues are prioritized as HIGH, MEDIUM, or LOW. - Security Scanning: Greps for hardcoded secrets, eval, dangerouslySetInnerHTML, string-concatenated SQL, and exposed API keys. - Use Case: After an AI agent finishes building a feature, run this Skill to verify every Spec item is truly implemented before handing fixes back to a dev-builder or bug-fixer agent. ## Quick Start Review my project code against Product-Spec.md and the design brief, then output a structured report with evidence for each conclusion.

Frequently Asked Questions about code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify code matches a product specification?▼

Check each Spec item against the codebase one by one, recording file and line evidence for every conclusion. This Skill automates that process, classifying each feature as fully implemented, partially implemented, or missing.

How to detect fake or misleading tests in a codebase?▼

Examine whether test preconditions match production reality, including units, reachable inputs, and assertion direction. Tests that only cover pure functions and happy paths are flagged as coverage blind spots.

What security issues does automated code review catch?▼

The review greps for hardcoded secrets, eval usage, dangerouslySetInnerHTML, string-concatenated SQL, absolute paths, and exposed API key patterns like sk-ant- or OPENAI_API_KEY. Suspicious patterns are verified via web search before being reported.

Does this code review require a design file to work?▼

No, only Product-Spec.md and the project code are required. Design briefs, design tool MCP integrations, Playwright, and git are optional enhancements that deepen UI consistency and visual comparison checks.

What happens after the review report is generated?▼

The report stops at findings; fixes are routed by the main agent. Stage 1 failures go back to a dev-builder for implementation, while quality issues go to dev-builder and defects or security issues go to a bug-fixer.