Content
78%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 high-quality, highly actionable instruction skill: every section gives concrete commands, paths, templates, and evidence requirements rather than vague direction, and validation steps are explicit. The main structural weakness is that it is a monolithic single-file manual — the long coalesce and dependency-swap procedures would work better as one-level-deep reference files, and a little redundancy could be trimmed.
Suggestions
Move the detailed 'Coalesce Superseded Agent Notes' procedure (the five-step chain plus the added-then-removed rules) and the 'Hand-Rolled Code Versus A Dependency' checklist into reference files, keeping a short summary with clearly signaled links in SKILL.md — this would raise progressive_disclosure from an inlined manual toward an overview-plus-references structure.
Deduplicate the dependency-policy material: it is summarized in 'What Counts As A Strong Candidate' and then restated as a full section; keep one pointer and one owner to trim tokens without losing content.
Add a brief explicit top-level sequence (e.g., a numbered 'Workflow' list naming the phase order: context → survey → prove/reject → write/inline → validate → PR) so the multi-phase flow does not rely solely on section ordering.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes competence — it never explains generic concepts — but there is redundancy that could be trimmed: the dependency policy appears in both "What Counts As A Strong Candidate" and the dedicated "Hand-Rolled Code Versus A Dependency" section, `knip` is qualified twice, and the thin-candidates paragraph overlaps the reject criteria. This fits anchor 4 (efficient with minor trimmable instances) better than anchor 5's every-token-earns-its-place, and better than anchor 3's "some unnecessary explanation". | 4 / 5 |
Actionability | Guidance is fully concrete and executable throughout: exact note paths (`.agents/notes/<lifecycle>/<class>/yyyy-mm-dd-topic.md`), exact validation commands (`pnpm run doc-sync`, `pnpm run lint`, `git diff --check`), concrete search patterns (`.name(` and `name(`), a full Agent Note section template, and copy-paste-ready tag examples like `TODO(double-default)`. For an instruction-only skill this matches the anchor-5 equivalent of copy-paste-ready commands covering the common cases. | 5 / 5 |
Workflow Clarity | Phases are clearly ordered (repo context → survey → prove/reject → coalesce → write → validate), the coalesce procedure is a numbered 5-step sequence with post-edit search verification, and validation commands plus a PR-summary checklist are explicit, so the destructive-operation cap does not apply. It sits at anchor 4 rather than 5 because the top-level cross-phase sequence is implicit in section order rather than explicitly enumerated with per-phase checkpoints. | 4 / 5 |
Progressive Disclosure | There are no bundle files (no references/, scripts/, or assets/), and while the body is well-sectioned with clearly signaled relative links to repo docs, it is a ~141-line full manual rather than an overview pointing to detailed materials. The dense "Coalesce Superseded Agent Notes" procedure and the dependency-swap procedure are reference-file material inlined in SKILL.md, matching anchor 3 (content that should be separate is inline) more than anchor 4. | 3 / 5 |
Total | 16 / 20 Passed |