Content
67%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 well-structured and actionable, with concrete templates, checklists, and right/wrong examples, plus explicit validation for destructive work. Its main weakness is conciseness: core ideas repeat across sections and some motivational prose could be trimmed.
Suggestions
Tighten conciseness by de-duplicating the 'three stops' (it appears in the opening, the core list, and its own section) into a single canonical statement that other sections reference.
Trim rhetorical/motivational sentences such as 'You are a senior operator, not an intern, and you do not need a babysitter' that restate the skill's premise without adding behavioral instruction.
Consider moving the four worked examples into a separate references file so the main SKILL.md stays a lean overview, which would also raise progressive_disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely efficient and avoids explaining basics Claude already knows, but the body (~170 lines) restates core themes multiple times ('the three stops' appears in the description, the core list, and a dedicated section) and includes motivational prose like 'You are a senior operator, not an intern'. Above 2 because it is not padded with background explanation; below 4 because several sections could be tightened without losing meaning. | 3 / 5 |
Actionability | Gives concrete, executable behavioral guidance: the BLOCKED/TRIED/NEED template, the four-point 'Proving it's done' checklist, the 'Assumed X; tell me if you meant otherwise' line, and right-vs-wrong examples. Below 5 because as an instruction-only skill some guidance is principle-level rather than copy-paste ready; above 3 because the templates and examples are directly usable. | 4 / 5 |
Workflow Clarity | 'Proving it's done' is an in-order numbered checklist with explicit validation (build/tests/linter, protected things, size match) and a fix-or-stop feedback loop; 'Before you send' adds a 3-check pass. Below 5 because some checkpoints are implicit rather than a strict validate->fix->retry loop; above 3 because validation checkpoints are clearly present for destructive/batch-style work. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/scripts/assets are empty), so the self-contained body relies on well-organized ## sections with no nested 2+ level references and nothing that clearly belongs in a separate file. Below 5 because at ~170 lines the body is longer than the 'under 50 lines' simple-skill case the top anchor describes; above 3 because structure and navigation are clear and no references are buried. | 4 / 5 |
Total | 15 / 20 Passed |