Content
75%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 well-built instruction-plus-scripts skill: runnable commands, explicit constraints handling, labelled research-vs-practice rules, and a clean overview-to-reference split. The main gaps are a broken reference to a missing references/papers/ directory, missing invocation examples for the planner and history scripts, and inline statistics that duplicate the evidence map despite the skill's own instruction to verify numbers in the paper files.
Suggestions
Fix or remove the 'references/papers/' pointer (and its INDEX.md mention) — the directory is absent from the bundle, so navigation dead-ends; either ship it or point paper verification at references/library.md and evidence-map.md.
Add a complete example invocation for strength_plan.py (e.g. the full command with --goal and athlete.json inputs) and one for strength_history.py, mirroring the copy-paste quality of the strength_check.py examples.
Trim the inline effect sizes and confidence intervals from the evidence section to direction-level statements, deferring exact numbers to references/evidence-map.md — this aligns with the skill's own 'open the file before quoting a sample size or effect' rule and cuts duplicated tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense and domain-specific with no general-knowledge padding, but the evidence section quotes full statistical detail (CIs, participant counts, SMDs) that duplicates references/evidence-map.md — and the Papers section itself instructs 'Before you quote a sample size, effect or limit, open the file', so the inline numbers could be trimmed to directions. | 4 / 5 |
Actionability | strength_check.py is copy-paste ready with flags, and planner inputs, user-choice keys, and the exit-2 stop-and-ask behavior are concretely specified. However there is no full invocation example for strength_plan.py's generation case or strength_history.py, and 'athlete.py show' is given without a path. | 4 / 5 |
Workflow Clarity | Explicit ordering ('Run athlete.py show first'), recovery-gated rules (RED/AMBER), a defined handoff output block, and stop-and-ask validation give most checkpoints. The end-to-end sequence across sibling skills must be assembled from the frontmatter description and scattered sections rather than one ordered workflow. | 4 / 5 |
Progressive Disclosure | Good one-level-deep structure: the ~60-exercise library and goal profiles live in JSON references, the evidence map in its own file, and the body is a genuine overview. But the body points to 'references/papers/' and its INDEX.md, which do not exist in the bundle — a broken navigation pointer. | 4 / 5 |
Total | 16 / 20 Passed |