Content
86%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 dense, highly actionable routing skill that maps user intents to exact CLI commands, includes validation and error-handling feedback loops, and defers detail to real one-level-deep references. It is slightly verbose where the runtime contract restates rules already implied by the routing and error sections.
Suggestions
Consolidate the repeated "do not guess/load the full catalog" guidance so it appears once (e.g. in 最小 DWS 执行契约) rather than being restated in Shortcut 发现, 按需加载, and 错误最短路径.
Consider moving the 12-bullet DWS runtime contract into a shared reference, keeping only the chat-specific deltas inline, to further trim the inline token footprint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and command-focused with no background concept explanation Claude already knows, but the runtime-contract and shortcut-discovery sections restate a few "do not guess/do not load" rules that slightly overlap the on-demand and error sections. | 4 / 5 |
Actionability | Provides copy-paste-ready commands with concrete arguments throughout (e.g. "dws chat +dm --to <姓名> --content <内容>", "dws chat group rename --id <openConversationId> --name <新名称>"), mapping intents to exact executable entries. | 5 / 5 |
Workflow Clarity | Clear sequencing via the Golden Route table plus explicit validation checkpoints ("查询结果必须检查 complete, hasMore, failures", partial results not reported as success) and an error-shortest-path section with retry/Trace-ID feedback loops; minor gaps as it is routing-oriented rather than one linear process. | 4 / 5 |
Progressive Disclosure | Clear overview with a well-signaled, one-level-deep reference table in "按需加载" whose paths (references/chat/message-query.md, references/card/create.md, etc.) are real bundle files, and an explicit "不要预加载 reference" directive. | 5 / 5 |
Total | 18 / 20 Passed |