Content
93%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 well-structured, token-efficient body: actionable bash examples, a decision table for command selection, an explicit candidate-confirmation checkpoint for side-effect flows, and clean one-level-deep references that all resolve to real files. The only modest gap is the absence of an explicit error-recovery feedback loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and efficient: a decision table plus short bash examples with no concept-padding (no explanation of what Feishu/open_id is); even the notes (41050, cross-tenant visibility, ID types) are Feishu-specific knowledge Claude does not already have, so every token earns its place. | 5 / 5 |
Actionability | Fully executable, copy-paste-ready commands covering common cases — 'lark-cli contact +search-user --query "张三" --has-chatted --as user', the full 'user_profiles batch_query' invocation with --params/--data, and search-bot variants — not pseudocode. | 5 / 5 |
Workflow Clarity | Clear sequence ('先搜,确认 open_id,再发') with an explicit validation checkpoint for side-effect operations ('搜索命中多条…把候选列给用户挑;不要擅自选第一条'), avoiding the batch/destructive cap of 3; falls short of 5 only because a full validate→fix→retry feedback loop is not spelled out. | 4 / 5 |
Progressive Disclosure | Clear overview (identity×task decision table) with well-signaled one-level-deep references to the three real bundle files (lark-contact-search-user.md, -search-bot.md, -get-user.md) and clearly marked out-of-scope pointers to sibling skills; detail is appropriately split into the reference files. | 5 / 5 |
Total | 19 / 20 Passed |