Content
77%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 content is highly actionable and the 6-phase pipeline is clearly sequenced with explicit clarification gates and failure fallbacks, supported by complete worked examples. The main weaknesses are token efficiency — duplicated trigger lists from the frontmatter and version-specific model recommendations that will age — and a monolithic single-file structure where the long example section could be offloaded to a reference file.
Suggestions
Move two of the three worked examples (e.g., Example 2 and 3) to a references/examples.md file and keep one representative example inline, leaving a clearly signaled one-level-deep reference so SKILL.md stays an overview.
Replace hardcoded model version names ("Sonnet 4.6", "Opus 4.6") with capability-based guidance (e.g., "use the strongest reasoning model for planning") or isolate them in a marked versionable section so they can be updated without touching the core instructions.
Remove the "When to Use" and "Do Not Use When" bullet lists that duplicate the frontmatter description verbatim, or compress them into a one-line pointer to reduce redundant tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — dense mapping tables instead of prose, and no explanations of concepts Claude already knows — but it can be tightened: the "When to Use" / "Do Not Use When" lists duplicate the frontmatter description almost verbatim, and time-sensitive version numbers ("Sonnet 4.6", "Opus 4.6", "Next.js 15", "Go 1.22" in the model-recommendation table and examples) are not isolated in a deprecated/versioned section, which the guidelines explicitly penalize. This sits between 'efficient with minor trims' (4) and 'could be tightened' (3), noticeably below the midpoint given the duplication and version-drift risk. | 3 / 5 |
Actionability | Guidance is fully executable for an instruction-only skill: exact detection heuristics ("package.json → Node.js / TypeScript"), intent and tech-stack mapping tables, a precisely specified 5-section output format with per-intent quick patterns, and three complete copy-paste-ready worked examples covering the common cases (vague Chinese prompt, moderate English prompt, EPIC project). This matches the anchor for copy-paste-ready output with examples covering common cases. | 5 / 5 |
Workflow Clarity | The instruction "Run this 6-phase pipeline sequentially" is backed by explicitly numbered Phase 0–5, each with concrete inputs and outputs. Validation checkpoints are explicit, not implicit: Phase 4's checklist with the gate "If 3+ critical items are missing, ask the user up to 3 clarification questions", Phase 0's failure fallback ("If no project files are found... flag 'tech stack unknown'"), and the guardrail for "just do it" / "直接做" requests. The skill is advisory-only, so the destructive/batch validation cap does not apply. | 5 / 5 |
Progressive Disclosure | The skill is a single ~380-line file with no references/, scripts/, or assets/ directories. Section headers are well organized, but the ~120-line Examples section and the large intent/tech-stack matching tables are inlined content that would fit one-level-deep reference files (e.g. references/examples.md), keeping SKILL.md as an overview. This matches 'some structure but content that should be separate is inline'; it is above a 2 because the operational tables are needed at runtime and navigation within the file is clear. | 3 / 5 |
Total | 16 / 20 Passed |