Content
92%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.
An exemplary lean body: copy-shaped commands, explicit authorization and failure-handling rules, and clean scope boundaries, with safety checkpoints (preview default, stop-if-missing, no auto-resend) woven in without verbosity. Only small gaps (no concrete example values, undocumented 'methods' output) keep actionability from full marks.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Around ten lines with zero padding: every sentence carries operational information (commands, flags, authorization rules, scope boundaries, failure handling). No concepts Claude already knows are explained, matching the 'lean and efficient' anchor exactly. | 5 / 5 |
Actionability | Concrete command syntax with placeholders, real flags ('--confirm', '--account'), and explicit stop conditions ('缺失时在请求前停止') — mostly executable guidance. Minor gaps keep it below anchor 5: no fully instantiated example (e.g. a real room ID / full string account ID format) and 'tiktok-live methods' is listed without any explanation of its output. | 4 / 5 |
Workflow Clarity | A simple skill whose single actions (send/like) are unambiguous, with explicit validation checkpoints built in: dry-run by default ('默认预览'), --confirm only with authorization, stop before the request when materials are missing, and verify-before-retry on timeout/disconnect ('超时/断连先核对，不自动重发或换通道补发'). This error-recovery guidance satisfies the checkpoint expectation even for external write operations. | 5 / 5 |
Progressive Disclosure | The skill is under 50 lines with no bundle files, and the body is well organized (commands block, parameters, scope/failure rules) — per the rubric's simple-skill exception this scores 5. Detail is appropriately delegated one level deep to tiktok-hunter and expert-osbi rather than inlined. | 5 / 5 |
Total | 19 / 20 Passed |