Content
82%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 an exceptionally dense, action-first facilitation playbook: concrete commands, precise behavioral thresholds, and a numbered Finalize sequence with genuine validation and re-validation loops. Its weaknesses are the readability cost of the compression — run-on sentences packing many clauses — and a monolithic main file where a few well-separated concerns (headless payload, HTML view rendering) could be pushed to the existing references structure.
Suggestions
Break the longest multi-clause sentences (the calling-agent parameter list in 'On activation', the render_view command and view-comment logic in 'Close') into short steps or a small list so the sequence survives a single reading.
Move the headless JSON payload specification and the Close/HTML-view rendering rules into reference files under references/, linked with one clear line each, to slim the main file further.
Add a brief explicit list of the shipped references (rubric-general.md, rubric-planning.md) and scripts (check_ids.py, render_view.py) near the top so a reader discovers the bundle without parsing the naming convention.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is maximally lean: every sentence carries operational instruction, no concept Claude already knows is explained, and there is zero padding, matching 'Lean and efficient; assumes Claude's competence; every token earns its place'. It is not 4 because there are no sections of over-explanation to trim — the only cost of the compression is readability, not token waste. | 5 / 5 |
Actionability | Concrete, copy-paste-ready commands appear throughout ('uv run {project-root}/_bmad/scripts/memlog.py append --workspace {run} --type <fact|decision|...> --text "..."', 'uv run {skill-root}/scripts/check_ids.py {run}/{id}-{slug}.md', 'count the body with wc -w') alongside precise behavioral rules ('No message asks more than two things', 'up to three' high-impact questions). It falls short of 5 due to minor gaps: the headless JSON payload keys are named but their exact shape is not specified, and the Close section's render_view invocation with its nested view/comment logic takes several readings to reconstruct. | 4 / 5 |
Workflow Clarity | The process is explicitly sequenced (activation 1-3, intents, facilitation paths, Finalize 1-9) with real validation checkpoints and feedback loops — 'fix every undefined, duplicate and withdrawn hit', 'rerun check_ids.py and fix every toc hit', 'Log each pass only once its edits are in' — matching 'Clear sequence with most checkpoints present'. It is not 5 because the extreme compression leaves some steps and conditions ambiguous (e.g., the calling-agent parameter paragraph and the template-extension resolution rules require re-reading to reconstruct the exact order of operations). | 4 / 5 |
Progressive Disclosure | Bundle references are real, one level deep, and clearly signaled: '{skill-root}/references/<name>.md' resolves to the shipped references/rubric-general.md and rubric-planning.md, and scripts/check_ids.py and scripts/render_view.py both exist and are invoked by exact path. It is not 5 because the single SKILL.md carries ~4,500 words with sections (the headless JSON payload spec, the Close/HTML-view rendering rules) that could live in reference files, and the rubric references are named only indirectly through a naming convention rather than an explicit listing. | 4 / 5 |
Total | 17 / 20 Passed |