Code Bug and Vulnerability Analyzer

Problem

When engineers ask a tool to review a code snippet, the issue is rarely only “is there a bug?” The same input may be a service module, a utility script, a frontend component, or a competitive programming solution. These cases have different priorities: engineering code needs checks for security, dependencies, edge handling, and maintainability, while contest code needs correctness, complexity, overflow, and judge-specific boundary behavior. A generic review template can produce vague advice.

How it works

The skill first performs code-type detection, then follows the matching analysis path:

• Engineering code: identifies language and context, checks logic defects, security issues, and code smells, and labels findings as Critical/High/Medium/Low. The output explains the impact, shows a before/after fix comparison, and provides copy-ready replacement code with the reason for each change.
• Contest code: identifies algorithmic patterns such as DFS, BFS, DP, binary search, greedy, and graph algorithms, analyzes time and space complexity, checks empty input, negative values, large inputs, integer overflow, division by zero, and out-of-bounds access, and suggests complexity or constant-factor optimizations.

Supported inputs include pasted snippets, source file uploads, and code screenshots. Screenshots use OCR to extract text, but low-quality images may be flagged as unreliable.

Boundaries

This skill is useful for local code review and algorithm diagnostics, but it does not replace full SAST, fuzzing, unit tests, or production audits. It can only analyze the context provided by the user and cannot infer external dependencies, database schemas, or runtime environments automatically. When confidence is low, it should state uncertainty instead of guessing.