Content
92%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.
An exceptionally tight, highly actionable writing skill: exact template, quantified limits, executable counting command, refresh loop, and a validation checklist. Its one real defect is that the single external reference (examples.md) is dangling — the file is absent from the bundle, so the large/small/non-UI example bodies it promises cannot be reached.
Suggestions
Ship the referenced examples.md alongside SKILL.md (with the large PR #16662-based example, a small one, and a non-UI one), or remove the reference and inline one compact example body in its place.
If the examples are deliberately kept out of the auto-loaded body, add a one-line inline miniature of the provenance section (a filled-in 2-row table) so a dangling or future-missing reference never leaves the reader without any worked instance.
Consider noting in the body where examples.md lives (e.g., a references/ directory) and keeping the link path consistent with the actual bundle layout so the reference can be verified mechanically.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: no explanation of what a PR is, a dense one-screen template, hard numeric limits ("at most 4 bullets of one sentence each", "12 words or fewer"), and only ~10 lines of rationale that double as decision rules ("Write only what serves those three", "Default to short"). Every token earns its place, matching the top anchor; score 4 would require over-explanation worth trimming, and the only rationale present is two short paragraphs that directly govern formatting choices. | 5 / 5 |
Actionability | Guidance is fully concrete and executable: an exact heading template with placeholders, quantified limits, a field-definitions table with inclusion/exclusion rules, and a copy-paste-ready bash/jq command for counting user turns from Claude Code session logs, including fallback behavior ("write `approx.` after each number") when no record exists. This matches the copy-paste-ready top anchor; score 4 would require missing key details, and even the tool-delegation case ("invent the equivalent for your own tool") is explicitly justified. | 5 / 5 |
Workflow Clarity | The sequence is clear and checkpointed: template → provenance definitions → counting from a record → keeping stats current on every edit (an explicit refresh loop: "Every edit to the body... recounts the table and rewrites the stamp") → a six-item pre-posting validation checklist covering each hard limit. This is a writing skill, not a destructive/batch operation, so no validation cap applies, and the checklist-plus-refresh-loop pattern matches the top anchor. | 5 / 5 |
Progressive Disclosure | The body itself is well organized into single-purpose sections of appropriate length, but its sole external reference — "See [examples.md](examples.md) for a large, a small, and a non-UI body" — points to a file that does not exist in the bundle (no references/ directory, no examples.md anywhere alongside SKILL.md). The promised size-calibration examples are therefore unreachable and are not inlined either. This lands at the 'some structure but references problematic' midpoint rather than 4, where references would be present and only organization would have minor gaps. | 3 / 5 |
Total | 18 / 20 Passed |