Content
96%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.
A tight, highly actionable skill body: executable commands with complete flag documentation, a section matrix, an ordered presentation workflow with explicit conditionals and fallbacks, and no filler. The only defect found is navigational: the referenced references/papers/ directory (and strength_history.py) are absent from the bundle, which keeps progressive disclosure at 4 rather than 5.
Suggestions
Ensure the referenced `references/papers/` directory (with `INDEX.md` and the study files) actually ships in the skill bundle, or correct the path — as delivered, a Claude following the Papers section's "open the file" instruction will find nothing there.
Give `strength_history.py` (and the sibling-skill tools like `athlete.py` and `paper_fetch.py`) an explicit skill-relative or <skills>-relative path, so the references resolve as reliably as the report.py command does.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — no explanations of concepts Claude already knows (no definition of HRV, Garmin, SVG, or heatmap). Every line carries skill-specific operational information: commands with all flags and defaults ("--date YYYY-MM-DD", "--format md|html|both (default both)", "--no-deep", "--no-adapt"), exact output paths, a section matrix as a compact table, and policy rules ("A meter is a distance to the goal, never a forecast"). This matches the 5 anchor ("every token earns its place") rather than the 4 anchor, which requires identifiable over-explanation to trim — none is apparent. | 5 / 5 |
Actionability | Fully executable, copy-paste-ready guidance: three concrete commands ("python -X utf8 \"<skills>/rampwise-report/scripts/report.py\" daily" etc.) with the <skills> placeholder defined up front, every option documented with its default, exact output locations ("~/rampwise-coach/reports/<date>-<mode>/"), and named follow-up commands ("athlete.py check-in", "athlete.py goal-add", "paper_fetch.py"). The common cases (daily/weekly/monthly, md/html/both) are all covered, matching the 5 anchor. | 5 / 5 |
Workflow Clarity | The workflow is clearly sequenced (Run → tell the user where the files are → read "What stands out" and the injury-risk watch first → pick 2-3 actionable findings → ask the listed questions and record answers → offer the next step) as an explicit numbered checklist. Checkpoints and feedback loops are present: "If the watch is AMBER or RED or readiness is RED, say so before any good news", "No goal set means no meters: say so and offer to add one", and the paper fallback "if it is missing or abstract only ... run paper_fetch.py ... and say what you are doing". The skill is non-destructive, so the validation cap does not apply; this fits the 5 anchor (explicit validation steps, feedback loops, checklist) rather than 4. | 5 / 5 |
Progressive Disclosure | Good structure overall: the body is an overview, scripts are invoked by command rather than inlined, and "Design notes (`references/design.md`)" is a clearly signaled one-level reference with a one-line inline summary so the file need not be opened. However, scored against the actual bundle: the Papers section directs Claude to "references/papers/" ("INDEX.md lists the read depth"), but that directory does not exist in the bundle (references/ contains only design.md), and "strength_history.py" is named without a resolvable path. These are minor organization gaps consistent with the 4 anchor ("most content appropriately placed; references mostly clear; minor organization gaps") rather than the 5 anchor, which requires all navigation to hold up. | 4 / 5 |
Total | 19 / 20 Passed |