Scientific Code Review

Use this skill when reviewing analysis scripts for correctness,
reproducibility, and adherence to scientific best practices.

Review Checklist

1. Correctness

• Do computations match the described methodology?
• Are array operations broadcasting correctly?
• Are edge cases handled (empty arrays, single samples, NaN propagation)?
• Are indexing and slicing operations correct (off-by-one errors)?
• Are statistical tests used with correct assumptions?

2. Reproducibility

• Are random seeds set for all stochastic operations?
• Are library versions pinned or documented?
• Can the script run end-to-end from raw data to final output?
• Are hardcoded paths replaced with arguments or config?
• Is the output deterministic given the same input?

3. Error Handling

• Are file I/O operations wrapped in try/except?
• Are user inputs validated before use?
• Are informative error messages provided?
• Does the script fail gracefully on bad data?

4. Code Quality

• Are functions small, focused, and well-named?
• Are magic numbers replaced with named constants?
• Is there adequate documentation (docstrings, inline comments)?
• Are imports organized (stdlib → third-party → local)?
• Is dead code removed?

5. Performance

• Are there unnecessary loops that could be vectorized?
• Is data loaded efficiently (chunked reading for large files)?
• Are intermediate results cached when reused?

6. Scientific Best Practices

• Is data integrity maintained (no accidental mutation of input data)?
• Are units tracked and documented?
• Are analysis parameters exposed as arguments, not buried in code?
• Are results validated against expected ranges?

7. Output & Reporting

• Are all outputs saved with meaningful filenames?
• Do figures include proper labels, units, and error bars?
• Is a session log or audit trail maintained?

Review Format

Present findings as a structured review:

## Code Review: [script_name.py]

### Summary
Overall assessment: APPROVE / REVISE / REJECT
Key concerns: [1-2 sentence summary]

### Issues
| # | Severity | Line(s) | Issue | Suggestion |
|---|----------|---------|-------|------------|

### Positive Aspects
- [Things done well]

### Recommendations
1. [Ordered by priority]

Severity Levels

• CRITICAL — Bug or scientific error that would produce wrong results
• WARNING — Could cause problems or reduces reproducibility
• STYLE — Code quality improvement, no impact on correctness
• INFO — Suggestion or best practice note

Important Guidelines

• Do not modify files or run code — review only.
• Do not review code without fully reading and understanding it.
• Do not suggest changes that would alter scientific conclusions
  without flagging the implications.

Domain Customization