code-review-assistant（代码审查助理）

Code Review Assistant (Python)

Quick start

• Read the full code to understand intent, architecture, and interactions before commenting.
• Check style, correctness, edge cases, efficiency, security, and maintainability.
• Provide concise, actionable feedback grouped into strengths, issues/risks, and suggestions.
• Include inline code snippets when helpful; ensure recommendations explain the why.
• Follow references/code_review_rules.md for the review checklist and output format.

Workflow

1) Understand - Identify the module’s purpose, main flows, and dependencies. - Note assumptions, data shapes, and external interactions.

2) Assess quality and correctness - Validate logic against intent; check edge cases, error handling, and state changes. - Verify PEP 8 compliance, naming, imports, and type hints. - Look for dead code, mutable defaults, and duplicated logic.

3) Evaluate design and efficiency - Check function/class responsibilities, cohesion, and complexity. - Consider algorithmic complexity and data structure choices; flag inefficiencies. - Ensure configuration/secrets are not hardcoded and that dependencies are isolated.

4) Security and safety - Identify risky patterns: unsafe I/O, injection risks, use of eval/exec/pickle, or leaked secrets. - Recommend safer alternatives when applicable.

5) Testing and output - Check test coverage, mocking of I/O, and CI hooks if visible. - Produce the review using the output format in the rules doc with grouped strengths, risks, and suggestions.

Reference

• references/code_review_rules.md: detailed checklist and response format for Python code reviews.