Аудит качества кода по всему репозиторию

engineering

code-quality-review-all is a Claude Code skill that systematically audits every evaluation module in a repository against a single code quality standard or topic. It runs the autolinter via `uv run python tools/run_autolint.py --all-evals --check`, supplements it with targeted grep or AST scripts when needed, and produces three artifacts under `agent_artefacts/code_quality/<topic_id>/`: a `README.md` documenting the topic and examples, a `results.json` with per-file findings keyed to relative paths and line numbers, and a `SUMMARY.MD` with statistics and remediation recommendations. On re-runs the skill reconciles new scan results with saved ones — dropping resolved issues, preserving `fix_status` on remaining entries, and appending newly found problems. It is the right choice for periodic consistency audits covering import patterns, pytest marks, formatting conventions, or any other repository-wide standard; use eval-quality-workflow instead when reviewing a single evaluation.