Review Code Quality

Produce an evidence-based review of the changed scope. Separate mechanical rule violations from heuristic design smells and avoid speculative findings.

Review workflow

1. Read ../using-power-of-ten/references/power-of-ten.md, ../using-power-of-ten/references/code-smells.md, ../applying-solid-principles/SKILL.md, and ../writing-maintainable-python/SKILL.md.
2. Read repository guidance and identify the requested change, affected public behavior, and changed lines. Review the diff first; open surrounding code only as needed to prove a finding.

3. Run the bundled checker from the repository root:

   python3 <skill-directory>/scripts/audit_python.py --git-diff

   For explicit files or directories, replace --git-diff with their paths. Add --docstring-style google|rest|any when the project does not use the NumPy default. The checker covers only high-signal, statically detectable cases; it does not replace the semantic pass.

4. Trace the ten rules in order. For Rules 3, 8, and 9, label conclusions as Python-profile findings rather than literal C-rule compliance.
5. Review PY001-PY005: import safety, explicit entry-point assembly, function responsibility, public annotations, and comprehension readability.
6. Review SOLID01-SOLID05 against demonstrated clients, extension points, substitution contracts, interface use, and dependency direction. Treat SOLID03 checker output as evidence, not proof of the other principles.
7. Review the touched design against all five smell families. Report a smell only when the code shows the catalog's signal and a concrete maintenance or correctness cost.
8. Check the documentation convention on changed public units (DOC01-DOC03). Read ../writing-docstrings/SKILL.md before reporting a docstring finding. A summary-only docstring satisfies every supported style and is not a finding on its own.
9. Run the repository's own formatter, linter, type checker, and focused tests. Do not substitute the bundled checker for project validation.
10. If asked to fix findings, make the smallest behavior-preserving change, rerun checks, and re-review the resulting diff.

Finding standard

Report only findings that are actionable and caused or exposed by the changed scope. Use:

[severity] ID — concise title
location: path:line
evidence: what the code demonstrably does
impact: concrete failure or maintenance cost
remedy: smallest safe correction

Use error for a clear rule violation or correctness risk, warning for a confirmed smell with material cost, and note for an explicit, justified tradeoff. Do not report style preferences, existing unrelated debt, or a smell name without evidence.

If no findings remain, say so and list the checks actually run. Never imply that this review proves safety or exhaustively detects every smell.

Suppressions

Honor # quality: ignore[ID] - rationale only when it is on the finding line or immediately above the relevant statement and contains a specific rationale. Challenge broad, stale, or circular rationales. Prefer correcting code over adding a suppression.