Автоматическое исправление проблем качества кода

engineering

code-quality-fix-all is a Claude Code skill that systematically resolves code quality issues identified by the code-quality-review-all skill, targeting any topic stored under agent_artefacts/code_quality/<topic>/. It begins by reading three files in the topic directory — README.md for fix patterns and examples, results.json for the full issue list, and SUMMARY.md for an overall picture — then classifies every fix as Easy (single-line), Medium (multi-line), or Hard (requires research). Before applying any changes, the skill interactively asks which issue types, evaluations, and complexity levels to include, plus an optional cap on the number of fixes per run. All edits land in a dedicated branch named agent/<short_description>, each modified file is validated for syntactic correctness immediately after editing, and the final phase runs the repository's linter (ruff, flake8, mypy, or similar) on every changed file. Teams dealing with accumulated technical debt in agentic codebases will find the workflow auditable and easy to rerun incrementally.