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 tight, highly actionable instruction skill: the roundtable lifecycle is sequenced end-to-end with exact formats, command semantics, and a complete org archive template, and the detailed original design is properly deferred to a single real reference file. The lone gap is in the archiving workflow, which has no validation or failure handling (e.g., missing ~/Context directory, failed write).
Suggestions
In the 存档 section, add a guard before writing — e.g., create ~/Context if it does not exist, and verify the file exists and is non-empty after writing before reporting the path to the user.
Add a brief fallback for the 讨论循环 command menu — e.g., how the host should respond if the user issues an unrecognized or out-of-scope instruction — so the loop's control flow is total rather than implicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line is skill-specific instruction — participant selection criteria ("几个人的立场要能互相较劲", "至少一位行外人"), the unified speech format with six action tags, the command menu semantics, the moderator's duties ("表面共识等于没讨论"), and the exact org file layout. Nothing explains concepts Claude already knows and no section is padded; this is an instruction-only skill where every token earns its place. | 5 / 5 |
Actionability | Guidance is fully executable: an exact speech template ("【人物名】【行动标签】：发言内容"), six enumerated tags, four commands with defined behavior, a concrete timestamp command (`date +%Y%m%dT%H%M%S`), a precise output path (`~/Context/{timestamp}--圆桌-{议题关键词}__roundtable.org`), and a complete org outline to copy. Per the rubric's code-vs-instruction note, absence of code is not penalized when guidance is this actionable. | 5 / 5 |
Workflow Clarity | The six-stage sequence (定议题 → 请人 → 开场 → 讨论循环 → 收场 → 存档) is clearly ordered, the loop's three steps (发言/综述/等指令) are explicit, and the per-round command menu plus the defined `止` exit make good checkpoints. However, the archiving step has no verification or error handling — no check that ~/Context exists or that the file was written successfully beyond reporting the path — so it falls short of the score-5 anchor's explicit validation and feedback loops. | 4 / 5 |
Progressive Disclosure | The body opens by pointing to `references/original-prompt.org` for the framework's original design (the host's persona and full process), a single clearly-signaled reference that exists in the bundle and itself contains no nested references. The remaining content — the operational loop, formats, and archive structure — is appropriately inlined for a skill of this scope, so structure matches the anchor for a clear overview with well-signaled one-level-deep references. | 5 / 5 |
Total | 19 / 20 Passed |