Content
88%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 highly actionable, well-sequenced coaching skill with strong validation gates and copy-paste templates; its main weakness is mild verbosity from repeated rules and the full inlining of material that could partly live in a one-level-deep reference file.
Suggestions
Dedupe the operating rules — 'Present, do not save', 'Always confirm before saving', and 'Never invent' each appear in both the 'Operating rules' section and 'Step 0'; state them once and reference them.
Consider extracting the App-Brief.md YAML/markdown template block into a references/brief-template.md file referenced one level deep, trimming the SKILL.md overview.
Trim the per-field challenge-quote examples (good examples kept in mind) to one representative line each to tighten token budget.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is largely efficient, assuming Claude's competence, but the recurring coaching-rule refrains (e.g. repeated 'Present, do not save' / 'Always confirm before saving' / 'Never invent' stated in both the rules section and Step 0) and lengthy challenge-quote examples could be trimmed without losing clarity. | 4 / 5 |
Actionability | Guidance is fully executable: exact pre-scan file-to-field mapping tables, a literal AskQuestion option structure, copy-paste YAML/Markdown templates for App-Brief.md, and concrete challenge prompts for each coached field with good examples to keep in mind. | 5 / 5 |
Workflow Clarity | A clearly sequenced Preflight → Step 0–5 flow with explicit validation checkpoints: the preflight skills-pull sentinel Glob, the 'do not write the file if any required field is still empty — return to coaching' gate, and feedback loops for retrying failed pulls and re-challenging vague answers. | 5 / 5 |
Progressive Disclosure | Content is well-organized into clearly signaled sections (Preflight, Operating rules, Step 0–5) with no nested references and no bundle files to mis-link; it scores 4 rather than 5 because substantial reference material (the per-field rubric, spec-kit mapping, and the App-Brief template) is fully inlined where a one-level-deep reference file could reduce the overview's bulk. | 4 / 5 |
Total | 18 / 20 Passed |