Bug Fix Assistant

What problem it addresses

In Python projects, many low-level bugs are not visible just by reading the code: date sorting without reverse=True, nested lists incorrectly using .append([...]), assignment written in an if x = y condition, missing colons, and bare except: blocks that swallow exceptions can all surface during runtime or code review. Manual search-and-replace across multiple files is easy to miss or overreach.

Bug Fix Assistant targets this class of deterministic, pattern-based Python issues and organizes scanning, fixing, and optional AI analysis into a library that agents can call directly.

How it works

It is not primarily a standalone CLI. It exposes bug_fixer.py for AI or automation workflows. Common entry points include:

• analyze(code, filename="input", use_ai=True): analyzes a code snippet, applies scans and auto-fixes, and optionally returns AI analysis.
• scan_file(filepath, auto_fix=True): scans a single .py file and writes back fixes by default.
• scan_directory(directory="."): scans Python source files under a directory while skipping .git, __pycache__, node_modules, virtual environments, and build directories.

The built-in rules fall into two groups. The first group consists of local deterministic rules that require no external API and run in milliseconds. Examples include adding reverse=True to date-related sorted(...) calls, changing nested-list .append([...]) to .extend([...]), correcting if x = y to if x == y, inserting missing colons, and replacing bare except: with except Exception:. Mutable default arguments such as def f(x=[]) are marked NEEDS-REVIEW rather than auto-fixed, to avoid changing runtime semantics.

The second group is AI deep analysis. When SILICONFLOW_API_KEY is configured, it can call SiliconFlow (DeepSeek-V3) to inspect complex business logic, potential null pointer or type issues, security risks such as SQL injection or XSS, performance and concurrency concerns, and code readability. If the key is missing, the network call fails, or a timeout occurs, the system degrades to deterministic rule results instead of aborting.

Boundaries

This skill is well suited for quickly cleaning frequent Python syntax misuse and low-level pattern bugs, but it is not a replacement for human code review or complex algorithm review. It only scans .py files and does not handle JS/Go or other languages. Automatic write-back overwrites the original file, so version control should be committed before bulk fixes. Very large files are read into memory, so they should be split or passed as snippets. Auto-fixes improve syntactic stability but do not guarantee business correctness, especially when sorting direction or comparison semantics are involved; developers should verify the results.