logic-lens – skill Lập trình cho agent AI

Logic Lens

Overview

Logic Lens is a Claude Code skill that performs deep, logic-driven code review using formal reasoning frameworks. Unlike traditional linters that check syntax and style, Logic Lens analyzes your code for logical errors, race conditions, security vulnerabilities, type mismatches, and algorithmic flaws that only appear when you reason through the code's behavior.

Powered by structured AI analysis, Logic Lens applies systematic logical inspection across 9 risk categories: null/undefined handling, type safety, concurrency, resource management, security injection, boundary conditions, algorithm correctness, state management, and API contract violations.

When to Use This Skill

• Use when you want a thorough logic review before merging a PR
• Use when a bug seems hard to find and standard linters aren't helping
• Use when reviewing security-sensitive code paths (auth, payments, file access)
• Use when refactoring complex business logic
• Use when onboarding to a new codebase and need to understand risk areas

How It Works

Logic Lens uses Claude Code's reasoning capabilities to:

1. Parse code structure and build a mental model of data flow

2. Apply formal logic checks across 9 risk categories

3. Trace execution paths for edge cases and boundary conditions

4. Identify security anti-patterns (injection, privilege escalation, data leakage)

5. Report findings with severity levels and actionable fix suggestions

Installation

# Install via Claude Code plugin marketplace
# Search: "logic-lens" in Claude Code > Extensions

# Or install via NPX (Antigravity)
npx agentic-awesome-skills --claude
# Then invoke: @logic-lens

Examples

Example 1: Review a Single File

@logic-lens review src/auth/login.ts for security issues

Logic Lens output:

[CRITICAL] SQL Injection risk at line 42: user input concatenated into query string
[HIGH] Missing rate limiting on login attempts
[MEDIUM] Password comparison uses == instead of timing-safe comparison
[LOW] Error messages may leak valid usernames (user enumeration)

Example 2: Full Repository Scan

@logic-lens scan the entire codebase and prioritize by severity

Example 3: Pre-PR Review

@logic-lens review all files changed in this branch before I open a PR

The 9 Risk Categories

| Category | What It Checks |

|----------|----------------|

| Null/Undefined | Missing null checks, optional chaining gaps |

| Type Safety | Implicit coercions, any-typed boundaries |

| Concurrency | Race conditions, shared mutable state |

| Resource Management | Unclosed handles, memory leaks |

| Security Injection | SQL/XSS/Command injection, path traversal |

| Boundary Conditions | Off-by-one errors, integer overflow |

| Algorithm Correctness | Wrong complexity, incorrect assumptions |

| State Management | Inconsistent state, missing rollbacks |

| API Contracts | Undocumented side effects, broken interfaces |

Best Practices

• Run @logic-lens on authentication and payment code before every release
• Combine with @lint-and-validate for full coverage: style + logic
• Review the CRITICAL and HIGH findings first; LOW findings can be deferred
• Use @logic-lens on legacy code you are about to modify to understand risk surface

Evidence and limits

The upstream contributor reports benchmark benefits, but this tree bundles no corpus, evaluation protocol or reproducible benchmark artifacts. Do not repeat a “47%” detection advantage as verified evidence. Produce concrete findings with source locations, reachable data/control flow, assumptions and a validating test or reproduction plan. An AI logic review is not a formal proof or complete security scan; an empty finding list is scoped absence of evidence.

Related Skills

• @lint-and-validate — Complementary: run after logic-lens for style/syntax
• @security-auditor — Specialized security-only deep scan
• @debugging-strategies — Use when logic-lens findings need tracing

Additional Resources

• [GitHub Repository](https://github.com/hyhmrright/logic-lens)
• [Dev.to Article: Why AI Code Review Misses the Most Dangerous Bugs](https://dev.to/hyhmrright/why-ai-code-review-misses-the-most-dangerous-bugs-logic-lens-fixes-that-4a8l)
• [Claude Code Skills Documentation](https://docs.anthropic.com/claude-code)

Limitations

Use this skill only when the task clearly matches the scope described above (code review and logic analysis). Logic Lens provides AI-powered analysis and should be combined with human review for production-critical decisions. Do not treat the output as a substitute for environment-specific testing or security audits.