What problem does it solve?
When you modify the logseq-review-workflow skill, it is hard to know whether the change actually improved review quality or introduced regressions. This Skill runs the same review prompt against isolated before and after snapshots of the skill, collects both outputs, and produces a structured comparison so you can judge the delta objectively.
Core Features & Use Cases
- Isolated Snapshot Setup: Extracts the old skill revision from a git ref and the new revision from a git ref or the working tree into separate snapshot directories, preventing cross-contamination between runs.
- Deterministic Output Comparison: Parses findings (severity, category, location, issue, impact, suggestion) from both raw outputs and reports shared, lost, and added findings plus severity and category counts.
- Qualitative Evaluation Rubric: Provides scoring criteria for finding quality, coverage, validation rigor, subagent orchestration, and final answer usability, with Improved/Regressed/Mixed/Inconclusive decision labels.
- Use Case: After editing routing rules in logseq-review-workflow, run this Skill with the previous commit as the before ref and the working tree as the after revision to verify the new revision keeps true findings and does not fabricate validation claims.
Quick Start
Use the logseq-review-workflow-eval skill to compare the before and after behavior of logseq-review-workflow on the same review prompt.
