Content
63%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is exceptionally concrete — real commands, flags, exit codes and validation loops — and the workflow is well sequenced with error recovery. It is dragged down by wall-of-text paragraphs, rules and options repeated across sections, study details inlined that already live in evidence-map.md, and several referenced files that do not exist in the bundle.
Suggestions
Break the 'Garmin workouts and the session menu' and heart-rate-rule sections into tables or lists, and strip the inlined study statistics (Lu 2025, Heiber 2024, Eur J Appl Physiol 2025) down to pointers into references/evidence-map.md, where they are already documented.
State the quality-session lock rules and the plan.py/build_plan.py --legacy relationship once each instead of three and two times respectively, and add the missing referenced files (references/papers/INDEX.md, docs/EVIDENCE-DEPTH.md) or remove the references.
Give the calendar connector operations and `athlete.py injury-add` the same explicit syntax treatment as build_plan.py so every instruction is copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Almost every token is skill-specific (settings paths, scripts, rules) with no explanation of concepts Claude already knows, but it is far from tight: the session-menu paragraph is a ~450-word run-on, the quality-lock rules are stated in three places, `--legacy` and the export formats each appear twice, and specific dates ("set by the athlete 2026-11-16") are inlined rather than isolated, so it sits at 'mostly efficient but could be tightened' rather than the minor-trim level of 4. | 3 / 5 |
Actionability | Concrete, mostly copy-paste-ready commands throughout: `python -X utf8 "<skills>/rampwise/scripts/athlete.py" show`, `build_plan.py --start YYYY-MM-DD --weeks 4 --readiness green|amber|red --out ...`, `athlete.py set hr.easy_band '{"run":[lo,hi],"bike":[lo,hi]}'`, `plan.py --list-events`; minor gaps keep it from 5 — the calendar connector operations (`list_events`, `delete_event`, `create_event`) and `athlete.py injury-add ...` are described in prose without their syntax. | 4 / 5 |
Workflow Clarity | A clear numbered sequence (State → Build → Explain → Calendar → Weekly re-sync) with explicit validation checkpoints (exit code 2 with the questions to ask, "every file is decoded again as a check", the event planner "stops and lists" missing inputs, list-before-delete-before-create on the calendar) and error-recovery loops; it falls short of 5 because the plan.py/build_plan.py legacy-forwarding relationship is explained late and confusingly, and cross-skill inputs (readiness, head-coach handoffs) are assumed rather than sequenced. | 4 / 5 |
Progressive Disclosure | References are one level deep and clearly signaled (`references/evidence-map.md` "full table", `library.md` "rows 20-24"), but the body inlines content that belongs in those references (Lu 2025 / Heiber 2024 sample sizes and bpm errors duplicated from evidence-map.md) and points to paths absent from the bundle (`references/papers/` with `INDEX.md`, `docs/EVIDENCE-DEPTH.md`), which is more than the 'minor organization gaps' of the 4 anchor. | 3 / 5 |
Total | 14 / 20 Passed |