Simplify

Simplify — three-pass code review

Run three passes over the target. Keep each pass narrow: don't let the reuse reviewer second-guess the quality reviewer, etc. After all three passes, synthesise — pick the smallest set of changes that improves the most dimensions.

Step 0: Locate the target

The user passed ${target}. Resolve it:

• If empty, default to the current diff vs the merge base of HEAD (or vs main / master when the merge base is hard to find):

  git diff --name-only origin/main...HEAD 2>/dev/null || git diff --name-only HEAD~1 2>/dev/null || git ls-files --modified
  

• If a path: read it directly (or, for a directory, glob for code files).
• If a git range (e.g. HEAD~3..HEAD, feature-branch...main): walk the diff.

If the target turns up empty, stop and ask the user what they'd like reviewed.

Step 1 — Reuse pass

Pass criterion: what is being re-implemented that already exists?

Look for:

• Utility functions duplicated across files. Same logic, different imports.
• Inline implementations that bypass an existing helper / library the project already depends on.
• Constants / enums redeclared per-call-site instead of imported once.
• Shape coercion / conversion code that an existing serializer already handles.

Output (this pass): bullet list of duplicated / re-implemented things, each with the existing canonical location + the new duplicate location.

Step 2 — Quality pass

Pass criterion: what would a reviewer flag in a PR comment?

Look for:

• Names that don't match what they hold (data, temp, info).
• Functions doing more than one thing (and / or in the function signature is a tell).
• Error handling that swallows information (except: pass, raw rethrow with no context).
• Comments that no longer describe the code.
• Dead code (unreachable branches, unused imports / params, no callers).
• Magic numbers that lack a named constant.

Output (this pass): bullet list, each item naming the file:line + the smell.