Content
71%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 well-executed, highly actionable instruction skill with a clear step sequence and thorough edge-case handling, held back by one internal inconsistency (the `content_hash` field referenced in delta classification but absent from the manifest example and scan instructions) and by packing both the status report and the full insights mode into a single ~305-line file with no reference files. Splitting insights into a bundled reference and resolving the hash field would lift the weakest dimensions.
Suggestions
Move the Insights Mode (the eight "What to compute" sections plus its full output template, ~120 lines) into a separate reference file such as references/insights.md, keeping only the trigger phrases and a short summary in SKILL.md so the main file stays an overview.
Resolve the content_hash inconsistency: either add a `content_hash` field to the manifest example and instruct computing it in Step 1's "Record:" lists, or drop the hash comparison and rely on the mtime/size comparison the manifest example actually shows.
Trim the large example blocks — the 40-line manifest JSON and the full status-report template — to only the fields and sections the skill actually reads or emits.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and procedural with no explanations of concepts Claude already knows, and it assumes competence throughout. Not 5: at ~305 lines for a read-and-report skill, the 40-line manifest example and two full output templates could be trimmed without losing guidance value. | 4 / 5 |
Actionability | Highly concrete for an instruction-only skill — exact glob patterns, decision thresholds ("<20% of total", ">50% of total"), the cohesion formula, wikilink scoring rules (+3/+2), and exact output templates. Not 5: Step 2 classifies sources by "hash differs from `content_hash`", yet the manifest example contains no `content_hash` field and Step 1's "Record:" lists never instruct computing a hash — a real executable gap. | 4 / 5 |
Workflow Clarity | Clear numbered sequence (Before You Start → Scan → Compute Delta → Report → Recommend) with explicit fallbacks for every failure mode (missing manifest, entries without `content_hash`, no previous snapshot, vaults under 20 pages, fresh rebuild). Not 5: the `content_hash` inconsistency leaves the core delta step ambiguous, and there are no verification checkpoints on the produced report. | 4 / 5 |
Progressive Disclosure | Headers are clear and the file is navigable, but there are no bundle files at all and two complete modes are inlined: the ~120-line Insights Mode with eight computation sections and a full output template clearly belongs in a separate reference file, with SKILL.md keeping a trigger summary. Fits "content that should be separate is inline"; not 2 because structure and navigation within the file are good. | 3 / 5 |
Total | 15 / 20 Passed |