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 a lean, command-driven reference for a single CLI tool: every section gives executable guidance, includes an explicit result-reauthorization requirement with exit-status validation, and documents real failure modes with concrete remediation commands. The main gaps are mild padding in the coaching section and an end-to-end workflow that must be assembled from multiple sections rather than read as one sequence.
Suggestions
Condense the 'Search coaching' section: keep the category-to-strategy mapping but drop the verbatim quoted replies, reducing it to one line per vague-query pattern.
Add a short ordered 'Workflow' section near the top (1. narrow vague queries, 2. `minutes search` with `--limit`, 3. reauthorize with `minutes get "<path>" --json` requiring exit 0) so the full sequence is readable in one place instead of being spread across Usage, Output, and Gotchas.
Consider moving the search-strategy tips and QMD setup details into a reference file (e.g., references/strategies.md) to keep SKILL.md focused on the core search-and-retrieve loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Efficient throughout — usage block, flags table, output schema, and gotchas are all command-first with no explanations of concepts Claude already knows. Not 5: the "Search coaching" section's verbatim suggested replies (e.g., the full quoted responses for broad queries) are somewhat padded and could be condensed to the coaching rule itself. | 4 / 5 |
Actionability | Copy-paste-ready commands everywhere: `minutes search "pricing strategy"`, `-t memo`, `--since 2026-03-01`, `minutes get "<exact path>" --json`, plus exact remediation commands (`qmd update && qmd embed`, `date -v-7d +%Y-%m-%d`). Common cases are covered with concrete examples; nothing is pseudocode. | 5 / 5 |
Workflow Clarity | The flow is clear with an explicit validation checkpoint ("require exit status 0") and error-recovery paths (empty results → `minutes list`; unindexed QMD → `qmd update && qmd embed`), and vague-query coaching acts as a pre-flight check. Not 5: the end-to-end sequence (search → coach → reauthorize via `minutes get`) is distributed across Usage, Output, and Gotchas rather than presented as one ordered workflow. | 4 / 5 |
Progressive Disclosure | No bundle files exist and the ~85-line body is well-organized into clearly labeled sections (Usage, Flags, Output, Gotchas) with nothing buried. Not 5: the skill exceeds the under-50-lines simple-skill threshold, and the Search coaching / Tips sections are arguably extractable reference material kept fully inline. | 4 / 5 |
Total | 17 / 20 Passed |