Content
78%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 body is a well-structured operational overview: lean, policy-dense guidance with concrete commands, flags, and data paths, a clearly sequenced task-based routing table, and a verified one-level-deep reference bundle that matches the links in the text. The only notable deductions are the inline version pin and minor duplication of acceptance-caveat sentences, plus the delegation of full command lines to the referenced install docs (appropriate for this design).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational with no explanations of concepts Claude already knows — every section carries routing rules, command flags, or data-path facts. It is not a 5 because of the inline version pin "目标运行包版本为 `0.1.4`" (time-sensitive information not placed in a deprecated/old-patterns section, which the guidelines penalize) and a few near-duplicate policy sentences such as "安装包验收不等于真实接入验收。Windows 说明是安装参考,不代表完整接入已验收。" | 4 / 5 |
Actionability | For an instruction-only skill the guidance is concrete: exact reference paths in a condition table, real command flags ("serve --web-only", "doctor --session-id"), env vars ("SN_PROACTIVE_AGENT_DATA_ROOT"), and precise data layouts ("projects/<project-id>/items/<item-id>/events.md", "runtime.jsonl"). It is not a 5 because full copy-paste command lines are delegated to the reference files, leaving minor gaps at this level; it is well above anchor 3 since nothing is pseudocode or vague. | 4 / 5 |
Workflow Clarity | The reading flow is explicitly sequenced ("先读通用流程,再按实际运行环境读取一份系统说明", read-only checks for status questions, one connector doc per harness), and validation semantics are stated concretely: "doctor --session-id 读取指定窗口的实际证据", "unverified 不等于通过", "安装包验收不等于真实接入验收". It is not a 5 because the step-by-step install/validation loop lives in overview.md rather than being fully spelled out here, leaving minor checkpoint gaps at the SKILL.md level; no destructive or batch operations are performed inline, so no cap applies. | 4 / 5 |
Progressive Disclosure | SKILL.md is a ~65-line overview whose condition table ("安装、运行与维护服务 | install/overview.md", "macOS | install/macos.md", etc.) signals exactly when to read each of the five bundle files, and all five referenced paths were verified to exist. References are one level deep — cross-links inside the reference files (e.g., overview.md → macos.md, hermes.md → ../install/overview.md) are sibling navigation, not nested 'see X for the real content' chains — matching anchor 5 (clear overview, well-signaled one-level-deep references, easy navigation). | 5 / 5 |
Total | 17 / 20 Passed |