improve(qa): rewrite as lean explore+report skill - #612
Conversation
- SKILL.md: 390 → 90 lines (77% token reduction) - Remove: browser tool, fix loop, bootstrap, tiers, regression mode - Add: exploration-guide.md reference (on-demand load) - Simplify: report template (fix sections removed) - Add: Phase 3 transition to subagent-driven/executing-plans
Prose paragraphs were too compressed for LLM guidance. Numbered lists with bold labels provide clearer step-by-step reference.
…ion Status skill_bytes: 3037 → 2702 (-11%). Compliance maintained at 95-100.
- Rules: 4 lines → 1 line (remove redundant "show evidence inline") - Fix: qa/references/ → references/ (match other skills' convention) - Phase 3: remove redundant If A/B/C lines (already in quote block) - skill_bytes: 2702 → 2533 (-6%)
skill_bytes: 2533 → 2502 (-1%). Compliance 88 (minor self-grading variance).
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (6)
📝 WalkthroughWalkthroughA QA skill redesign shifts scope from iterative fix-and-test cycles to a streamlined analyze-and-report model. The changes include a new exploration guide, simplified report template, trimmed issue taxonomy, and rewritten skill definition spanning three phases: analyze project characteristics, explore per strategy while recording issues, and transition to fix options or completion. Changes
Sequence DiagramsequenceDiagram
participant User
participant Analyzer
participant Explorer
participant Reporter
participant Transition
participant Subskill
User->>Analyzer: Invoke QA skill
Analyzer->>Analyzer: Analyze project structure,<br/>entry points, test presence
Analyzer->>Analyzer: Establish QA strategy<br/>from characteristics
Analyzer-->>Explorer: Strategy determined
Explorer->>Explorer: Iterate through<br/>exploration checklist
Note over Explorer: Per project type:<br/>(Web/CLI/API/Library)
Explorer->>Explorer: Reproduce each issue<br/>with evidence
Explorer->>Reporter: Collected issues & evidence
Reporter->>Reporter: Generate qa-report-{date}.md<br/>with evidence links
Reporter->>Reporter: Compute health score<br/>& write baseline.json
Reporter-->>Transition: Report ready
Transition->>User: Prompt: Proceed with fixes?
User->>Transition: Choose option:<br/>Subagent-driven OR<br/>Inline execution OR End
alt Proceed with fixes
Transition->>Subskill: Invoke subagent-driven-development<br/>OR executing-plans
Subskill-->>User: Fixes applied
else End
Transition-->>User: QA complete, no fixes
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Poem
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Summary
Rewrite
/qaskill from a 398-line monolithic explore+fix flow to a lean 65-line explore+report skill with on-demand references. 71% token reduction per invocation while maintaining 88-100% agent compliance across 5 autoresearch experiments.Changes
qa/references/→references/to match other skills' conventionTests
Summary by CodeRabbit