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.
A tight, actionable skill body: executable commands, exact output templates, and unusually good anti-hallucination gotchas. It loses points on small coherence issues — contradictory staleness defaults (14 vs. 7), an orphaned `minutes commitments --json` bullet with no workflow context, and a slightly oversized body whose formatting templates could move to a reference file.
Suggestions
Reconcile the staleness threshold: the main command runs `--stale-after-days 14` while the filter list says the default is 7 — state one default and explain why the example overrides it (or drop the flag from the example command).
Give the "Relationship-graph commitments come from `minutes commitments --json`" bullet a home: either add a step that runs it when stale commitments are found, or remove the bullet since `minutes commitments` appears nowhere else in the workflow.
Move the three report-format templates (decision conflicts, resolved supersessions, stale commitments) into a `references/report-templates.md` and link to it, keeping the SKILL.md body under 50 lines and earning the simple-skill progressive-disclosure exception.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence: a copy-paste command, filter flags, output templates, and genuinely novel gotchas ("The consistency check reads live Markdown... do not suggest deleting or rebuilding `graph.db` (it does not exist)") with zero concept explanations. Every section adds information Claude could not infer. | 5 / 5 |
Actionability | Provides an executable command (`minutes consistency --stale-after-days 14`), concrete filter flags, and exact report templates covering conflicts, resolved supersessions, stale items, and the clean case — matching the minor-gaps anchor. Not 5: the JSON output fields are never shown (only inferred from templates), and the staleness default is contradictory — the main command passes `--stale-after-days 14` while the filter docs say "default: 7" — leaving the reader unsure which threshold applies. | 4 / 5 |
Workflow Clarity | The run → parse → format → suggest-next-steps sequence is clear, with an explicit error-handling rule ("treat a nonzero exit as unavailable, never empty") covering the failure checkpoint. Not 5: the bullet "Relationship-graph commitments come from `minutes commitments --json`" sits under "When to suggest next steps" but references a command never invoked anywhere in the workflow, creating a gap in an otherwise coherent sequence; not 3 because the read-only lint workflow needs no further validation checkpoints. | 4 / 5 |
Progressive Disclosure | Well-organized single-level sections (How to run, Formatting the report, When to suggest next steps, Gotchas) with no nested references and appropriately inline content — matching the good-structure anchor. Not 5: the body is ~70 lines, above the under-50-line simple-skill threshold, and the per-section report templates are the kind of content that could split into a one-level reference file. | 4 / 5 |
Total | 17 / 20 Passed |