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.
A well-structured, highly actionable wizard with a clear staged workflow, real validation checkpoints, and clean progressive disclosure to three real reference files. The only soft spot is a handful of coaching sentences that pad the otherwise lean prose.
Suggestions
Trim soft coaching lines (e.g. '角色是用户的，你只负责把他想清楚的东西落到纸上，不替他做主') that restate the conversational principle already captured in 总原则, to tighten conciseness toward a clean 5.
Collapse the prose preamble of 阶段〇 ('用户通常只有一句话，比如…') into the single actionable instruction '先复述你理解到的火花，确认大方向，再进入采访' to cut tokens without losing guidance.
Move the meta-commentary in 阶段一 ('AI 天生爱写正确的纸纸人，缺陷是角色活过来的关键') into references/anti-slop.md where the anti-slop rationale already belongs, keeping the body purely procedural.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly lean and operational with concrete commands, paths, and templates rather than concept explanations Claude already knows, but a few coaching sentences (e.g. '用户通常只有一句话', '角色是用户的，你只负责把他想清楚的东西落到纸上，不替他做主') could be trimmed — efficient with minor over-explanation, not a clean 5. | 4 / 5 |
Actionability | Highly actionable with concrete commands and endpoints — 'web_search site:clawhub.ai', 'install_skill 的 github_url', 'ls ~/.hanako/skills', 'POST /api/character-cards/plan', 'POST /api/character-cards/import' — plus file paths and copy-oriented identity.md/ishiki.md/description.md templates covering common cases. | 5 / 5 |
Workflow Clarity | Clear eight-stage sequence (阶段〇–七) with explicit validation checkpoints: '每个阶段产出可见的东西，给用户确认后再往下走', a plan→confirm→import feedback loop, a 隐私红线 checklist, and an error-recovery section (常见状况) addressing mid-flow changes, id collisions, and zip-only delivery. | 5 / 5 |
Progressive Disclosure | Clear overview body with well-signaled one-level-deep references to real files — '先读 references/character-craft.md', '必读 references/anti-slop.md', '必读 references/card-format.md' — with heavy detail (card.json format, zip/API details, question bank) appropriately split into those references. | 5 / 5 |
Total | 19 / 20 Passed |