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 solid, actionable guidance skill: concrete executable bash at every step, a clearly sequenced doc-location fallback chain, and honest output-quality guardrails (no fabrication, state when info comes from the official site, flag version differences). The main improvements are small correctness details in the copy-paste code and splitting the fallback heuristics into a bundled script.
Suggestions
Fix the find expression to parenthesize the -o branch — `find . -type f \( -name "faq.en.md" -o -name "config.zh.md" \)` — so both patterns are actually matched.
Add explicit emptiness checks after each fallback find (e.g., `[ -n "$DOCS_DIR" ] && [ -d "$DOCS_DIR" ]`) so the prose checkpoints are enforced in the code, strengthening workflow validation.
Move the root-derivation and global-fuzzy-search heuristics into a scripts/locate_docs.sh bundle file, keeping SKILL.md to the decision flow — this trims the body and improves progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly lean, bash-driven guidance with minimal concept explanation, but slightly padded: the memory-lookup step is narrated twice ("首先你可以查看memory中是否有文档目录…如果没有则继续执行下一步" and again after the code block) and some comments restate obvious commands ("# 使用 dirname 获取该文件所在的目录"). | 4 / 5 |
Actionability | Provides concrete, executable commands throughout (python DOCS_DIR import, which/sed root derivation, find fallbacks, memory format template). Minor gaps keep it from copy-paste-perfect: `find . -type f -name "faq.en.md" -o -name "config.zh.md"` lacks parentheses so -type f binds only to the first predicate, and the memory-lookup find result is used without an explicit emptiness check. | 4 / 5 |
Workflow Clarity | A clear five-step sequence with an explicit fallback chain and existence checks ("如果上面获取到了非空路径且目录存在，直接使用"). Not 5 because several checkpoints live only in prose rather than the code (e.g., empty find results are not explicitly tested), so the validate-fix-retry loop is implicit in places. | 4 / 5 |
Progressive Disclosure | A well-organized single-file skill (no references/scripts/assets exist) with clear step headers and no broken or buried references. The ~135-line body inlines the fallback search heuristics and memory-record boilerplate that could live in a scripts/ or references/ file, leaving minor organization gaps. | 4 / 5 |
Total | 16 / 20 Passed |