Content
82%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 lean, highly actionable CLI skill document with concrete commands, hard limits, confirmation gating, and explicit error-handling rules including what remains unimplemented. Main gaps: the multi-step workflow is prose-scattered rather than sequenced, and the bundled scripts are never referenced from the body.
Suggestions
Present the core flow as a short ordered list (1. create 2. info 3. send 4. listen) so the sequence and checkpoints are explicit in one place.
Mention the bundle scripts (e.g. scripts/douyin_im.py, scripts/im_wire.py) and how the douyin-im CLI relates to them, so navigation to the bundle is possible.
Trim the Relay/封包/context-signing internals, keeping only the operative rule: the script handles signing; never reuse a context.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational: every line carries skill-specific facts (session file location and 0600 permissions, "单条文本 ≤1000 字，附件 ≤10MB", preview-by-default semantics, unimplemented features). It falls short of anchor 5 because a few sentences explain protocol internals the script already handles automatically, e.g. "创建会话先向 Relay 请求封包扩展，再以最终封包和一次性 context 签请求；不得重用 context" — trimmable detail for the agent. | 4 / 5 |
Actionability | The two command blocks are copy-paste-ready CLI invocations covering all common cases: create, info, send text/image/video/file/card-video, listen, and send-to, each with concrete flags, absolute-path placeholders, and limits ("--duration 60 --max-events 100 --output /受控目录/messages.jsonl"). The follow-up prose adds executable rules (--confirm gating, --content-file usage) rather than vague direction. | 5 / 5 |
Workflow Clarity | The sequence is stated ("创建成功后复用返回的 conversation_file，后续先用 info 查询，不重复创建") with real validation checkpoints: preview-by-default before sending, "操作前核对目标", and an explicit error-recovery loop ("请求结果未知时先查会话并报告，不自动重发私信", "PLATFORM_IM_INVALID_RESPONSE 表示响应无法解析，不代表票据失效"). Not anchor 5 because the multi-step flow (create → info → send → listen) is scattered across paragraphs rather than presented as one clearly ordered checklist. | 4 / 5 |
Progressive Disclosure | At ~37 lines with well-organized sections, the body appropriately keeps everything inline per the simple-skill guideline. However, the bundle contains scripts/douyin_im.py and scripts/im_wire.py and the body never references either — the "douyin-im" CLI's relationship to the shipped scripts is unexplained, a minor organization/navigation gap against the actual bundle structure. | 4 / 5 |
Total | 17 / 20 Passed |