Content
100%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 exemplary instruction-only skill body: a tight three-part structure (Ask, Then, Rules) with concrete file paths, exact gitignore entries, a specified commit message, and validation safeguards at every risky step. Template detail is correctly deferred to the bundle files rather than inlined.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean throughout ('Two layers and a contract, nothing else'), explains nothing Claude already knows, and every line in Ask/Then/Rules carries an instruction — matching the 'every token earns its place' anchor with nothing to trim. | 5 / 5 |
Actionability | For an instruction-only skill the guidance is fully executable: exact directory list 'wiki/{sources,entities,concepts,comparisons,synthesis,questions}', exact .gitignore entries, a copy-paste commit message 'kb: bootstrap <name>', and explicit prohibitions like 'Never use git add .'. | 5 / 5 |
Workflow Clarity | The 7-step sequence embeds explicit validation checkpoints (inspect a non-empty target and ask before adopting, inspect git status and the staged diff, check the diff for secrets before committing) plus an error-recovery path for failed commits, so the destructive/batch cap at 3 does not apply. | 5 / 5 |
Progressive Disclosure | Well-organized sections with one-level-deep references to references/templates/; all five referenced template files (CLAUDE.md.tmpl, index.md.tmpl, log.md.tmpl, overview.md.tmpl, contradictions.md.tmpl) were verified to exist, and no template bulk is inlined into the body. | 5 / 5 |
Total | 20 / 20 Passed |