dual-axis-skill-reviewer

Dual-axis scoring for automated code review combines deterministic code-based checks for structure and execution safety with LLM deep review findings to produce a reproducible quality score. This method evaluates scripts and tests to ensure comprehensive quality assurance.

To automate skill review and gate merges, run the command-line review script to generate a reproducible quality score. You can then enforce a score threshold, such as requiring a minimum of 90, to block low-scoring skills from merging into the project.

Yes, you need Python and an LLM integration to run automated skill reviews. The execution and scoring process specifically requires the uv package manager to run the review script and an LLM to generate the deep review findings.

Yes, you can run cross-project skill reviews from the command line. The reviewer script accepts a project root directory argument, allowing you to target and evaluate different projects individually to generate isolated quality reports.

The best way to identify improvement items for low-scoring skills is to use a dual-axis review approach. The LLM deep review evaluates the code and provides concrete improvement items directly alongside the deterministic quality score.

Deterministic checks evaluated during an automated skill review include project structure, scripts, tests, and execution safety. These code-based checks form one axis of the scoring system before being combined with LLM deep review findings.