A/B test CLAUDE.md instruction changes against eval benchmarks. Capture baselines, test variants, compare results.
The English summary is being prepared. Showing the original description.
See how to installReview the instructions the agent will receive before you install.
Systematically improve Mycelium instructions through measurement. Adapted from n-trax.
baseline -- Capture current performance/mycelium:eval-runner run-split optimization — record as optimization scores/mycelium:eval-runner run-split holdout — record as holdout scores.claude/optimization/baseline.json: timestamp, CLAUDE.md hash, optimization metrics, holdout metrics, overall and per-category metricstest <variant> -- Test a variant.claude/optimization/variants/<variant>.md/mycelium:eval-runner run-split optimization — this is the hill-climbing signal/mycelium:eval-runner run-split holdout — this validates generalization.claude/optimization/results/<variant>.jsonreport -- Compare all variantsGenerate comparison table with split-aware columns:
| Variant | Opt Pass Rate | Holdout Pass Rate | Delta Opt | Delta Holdout | Overfit? | Decision |
Flag Overfit? = YES when optimization delta is positive but holdout delta is negative.
exemplar <eval-name> -- Capture winning trajectoryAfter a clean eval win (1 iteration, fast), save the approach to .claude/optimization/exemplars/.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Accessibility audit, scoped to the surfaces a product actually has. Detects web / rendered_markdown / terminal / native_app / video_audio / document / headless, then applies only the criteria that bind. WCAG 2.1 AA in full for web; not at all for headless.
The English summary is being prepared. Showing the original description.
Bring Mycelium into a project that already has code. Detects that the repo predates the framework, asks before touching anything, then reads the codebase to draft what it CAN establish (delivery, solution shape) and — the actual point — names what it cannot (purpose, strategy, real user evidence). The output is a discovery backlog with a head start, never a filled canvas.
The English summary is being prepared. Showing the original description.
Design the smallest viable test to validate or invalidate a critical assumption. Based on Torres's assumption testing framework, organized by Gilad's AFTER model (Assessment → Fact-Finding → Tests → Experiments → Release Results).
The English summary is being prepared. Showing the original description.
Use before any research activity or significant decision. Reviews cognitive biases relevant to the current stage.
The English summary is being prepared. Showing the original description.
Use to evaluate whether current work aligns with Better Value Sooner Safer Happier. Run at diamond completion and periodically.
The English summary is being prepared. Showing the original description.
Lint canvas files for staleness, missing fields, inconsistent evidence types, and orphaned references. Run periodically or before major transitions.
The English summary is being prepared. Showing the original description.