Content
56%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 as a progressive-disclosure entry point with a clear RED/GREEN procedure, an explicit validation gate, and read-when cues on every reference. Weaknesses are moderate: redundant/meta sections, procedure steps that lack executable detail, and references that all point outside the skill bundle via non-resolvable repo: paths.
Suggestions
Merge the 'Use' and 'When to use' sections and drop the meta-commentary about archived vs. deprecated context and the 'Philosophy' section to tighten token efficiency.
Add one concrete executable example, such as a sample RED test and the repository test command pattern, to make the procedure copy-paste actionable.
Ship the referenced guides (SKILL.full.md, tests.md, mocking.md, interface-design.md, deep-modules.md, refactoring.md) inside the skill's own references/ directory with standard relative links, and remove the duplicate subagent-routing reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient, but includes unnecessary material: meta-commentary about its own architecture ('Archived means active deferred context, not deprecated guidance'), a redundant pair of 'Use' and 'When to use' sections, and a one-line 'Philosophy' section. Not 4: the redundancy across sections is more than minor; not 2: it never explains concepts Claude already knows (no library tutorials or basic TDD primers). | 3 / 5 |
Actionability | Some concrete guidance exists ('Produce a failing test first (RED), then apply the smallest fix (GREEN)' and the executable 'bin/ask skills audit ... --json' command), but the procedure is otherwise high-level with no example test command or concrete illustration of translating reproduction steps into a RED test. Not 4: key executable details are missing; not 2: the validation command and explicit inline-fallback routing exceed high-level hints. | 3 / 5 |
Workflow Clarity | A clear 5-step numbered Procedure with an explicit validation checkpoint ('Fail fast: stop at the first failed gate and do not proceed') and a hard constraint ('Do not skip RED verification'). Not 5: there is no fix-and-revalidate feedback loop or checklist describing what to do after a failed gate; not 3: the validation checkpoint is explicit, not implicit. | 4 / 5 |
Progressive Disclosure | References are one level deep and well-signaled with per-reference 'Read when:' cues, but the skill's own bundle contains no referenced files at all — every link points outside the skill directory via a non-standard 'repo:' link scheme, the subagent-routing reference is duplicated across two sections, and none of the paths resolve within the skill. Not 4: out-of-bundle, non-resolvable references and duplication are more than minor organization gaps; not 2: references are clearly signaled and organized, not buried. | 3 / 5 |
Total | 13 / 20 Passed |