review

SKILL.md

Code Review

Review the provided code (file, diff, or selection) systematically.

Review Checklist

  1. Correctness — Logic errors, off-by-ones, edge cases, wrong assumptions
  2. Bugs — Null/None handling, uninitialized variables, race conditions
  3. Types & shapes — Tensor dtype/device mismatches, shape errors, implicit casts
  4. Security — Injection, hardcoded secrets, unsafe deserialization
  5. Style — Naming, dead code, unnecessary complexity, consistency with surrounding code
  6. Performance — Unnecessary copies, repeated computation, memory leaks

Output Format

For each issue found:

  • File and line: where the issue is
  • Severity: critical / warning / nit
  • What: one-line description
  • Why: brief explanation
  • Fix: suggested change (code snippet if helpful)

If no issues are found, say so — don't invent problems.

Scope

$ARGUMENTS

Weekly Installs
3
GitHub Stars
7
First Seen
9 days ago
Installed on
cline3
github-copilot3
codex3
kimi-cli3
gemini-cli3
cursor3