Content
85%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 highly actionable, well-sequenced skill with real validation checkpoints and a clean, correctly-wired bundle script. Its one real weakness is token efficiency: the 'speed is the feature' rationale and reassurance asides are restated multiple times and could be cut roughly in half without losing any operational content.
Suggestions
State the speed rule once (e.g., in the opening or Gotchas) and delete the repetitions — 'Don't ask which one — that defeats the speed promise', 'That's a second prompt and it breaks the speed promise', and the Gotchas restatement of the two-question limit all restate the same constraint.
Trim reassurance asides with no operational content ('Don't be precious about it', 'Both are valid', 'That's fine', 'don't lecture them, just let them know') to single clauses or remove them.
The Phase 5 nudge quote and marker-file mechanics could be shortened to the marker-check command plus one sentence, since the marker file itself is already described as the entire state mechanism.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Operational content (commands, tag definitions, script contract) is efficient, but the speed rationale is repeated at least four times ('Tagging should take 5 seconds, not 5 questions', 'that defeats the speed promise', 'If tagging takes more than two questions ... you've broken it', 'it breaks the speed promise') and conversational filler like 'Don't be precious about it', 'Both are valid', and 'That's fine' pads the text. Mostly efficient with some unnecessary explanation — anchor 3, not anchor 4's 'minor instances'. | 3 / 5 |
Actionability | Fully executable throughout: exact `minutes list/search/get` commands per pattern, a copy-paste-ready `python3 .../tag_apply.py` invocation with argument specification, the complete AskUserQuestion option set with definitions, worked note-parsing examples, the one-time nudge marker commands, and a justified Edit-tool fallback. Specific examples cover the common cases, matching the anchor-5 example. | 5 / 5 |
Workflow Clarity | Five clearly sequenced phases with explicit validation and feedback loops: Phase 5 verifies parseability via `minutes get ... | head -3` with defined error handling ('If the output contains an error or warning about malformed frontmatter, surface it gently'), the script contract specifies `{"status": "ok"}` / `{"error": "..."}` outcomes with non-zero exit, and edge cases (idempotent re-tagging, atomic writes, missing frontmatter) are all addressed. This is a single-file, non-destructive edit with atomic-write protection, so the destructive/batch cap does not apply. | 5 / 5 |
Progressive Disclosure | Scored against the actual bundle: one bundled script (`scripts/tag_apply.py`) referenced via the correct `${CLAUDE_PLUGIN_ROOT}/skills/minutes-tag/scripts/tag_apply.py` path, one level deep, with its behavioral contract summarized in the body rather than duplicated. All inline content (phases, gotchas) is operational guidance that belongs in SKILL.md — nothing reads as reference material that should be split out — and section headers make navigation easy. | 5 / 5 |
Total | 18 / 20 Passed |