Content
75%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 tight, well-organized editing-conventions skill: actionable rules, a clear WHAT+WHEN frontmatter anatomy, and an explicit post-edit checklist plus validation procedure. Slight verbosity from restating the upstream best-practices highlights and repeating the 1024-char ceiling, and no on-disk bundle to verify references against.
Suggestions
Trim the restated Anthropic 'Canonical guidelines' highlights (Hard limits, Body size, Description structure) since they duplicate the upstream link; keep only the repo-local reinforcement.
Consolidate the 1024-char ceiling guidance into one place; it currently appears in both 'Token economy' and 'When to refactor a description'.
Add a concrete runnable command for measuring the folded description length (e.g. a one-line awk/python snippet) to make the validation step copy-paste ready rather than prose.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean rules and decisions (markdown style, frontmatter, token economy, progressive disclosure) with a few slightly over-explained passages (the restated em-dash ban, repeated 1024-char reminders) that could be trimmed, but it generally assumes Claude's competence and does not pad with basics. | 4 / 5 |
Actionability | Concrete, executable guidance dominates: 'Wrap markdown at 100 characters', 'Use the YAML folded block scalar (`>-`)', a copy-pasteable frontmatter example, and an explicit measurement procedure ('count the length of the single string YAML produces ... confirm it is under 1024'). Minor gaps: the validation step relies on prose rather than a runnable command. | 4 / 5 |
Workflow Clarity | Multi-step processes are well sequenced: the 'After Editing a SKILL.md Body' section lists Add/Remove/Tighten/State-no-change as an explicit checklist, and 'Validating a skill change' gives an ordered exercise-the-procedure workflow with an explicit checkpoint (confirm the skill loads and the agent follows it). The destructive/batch cap is not triggered since this is an editing-conventions skill, though it lacks an explicit error-recovery feedback loop. | 4 / 5 |
Progressive Disclosure | Content is organized into clear topical sections (Markdown style, Frontmatter, SKILL.md vs references, Canonical guidelines, Third-party skills, Validating) with the Anthropic upstream spec pushed to an external link (one level deep, clearly signaled). No bundle files are present, and the body is appropriately self-contained for this meta-skill, though some canonical-guideline highlights restate upstream material that could live in a reference. | 4 / 5 |
Total | 16 / 20 Passed |