← ClaudeAtlas

feynman-reviewlisted

Feynman review for science-explainer content: nitpick as a zero-background reader across six dimensions plus scoring; failures are sent back. Mandatory review for all science-explainer deliverables. 中文:科普内容费曼审查:假装零基础读者挑刺,六大维度+评分,不合格打回。所有科普类产物强制过审。
Culeot/content-forge-kit · ★ 1 · Code & Development · score 60
Install: claude install-skill Culeot/content-forge-kit
# Feynman Review Playbook User requirement: from now on, **all science-explainer content** (Xiaohongshu posts, video scripts, knowledge explainers) must pass this "zero-background reader" review after generation. This is a long-term mechanism, not a one-off action. ## Review Process 1. **Enter the zero-background state**: pretend you know nothing about the field and understand only from the content itself. 2. **First read**: write down "what did I learn" — how many points can you retell? 3. **Second read**: note "what would I ask" — where does it feel off, which term is unclear, which number is left hanging. 4. **Check against the original**: answer each of your own questions one by one, and find where the content failed to meet the reader's expectations. 5. **Score and conclude** (see below). ## Six Review Dimensions 1. **Phenomenon chain**: can a layperson retell the full process? (skipped steps / scrambled order → fail) 2. **Principle hook**: is the biggest counterintuitive point explained? Is what readers most want to know covered first? 3. **Terms grounded**: is every term defined in one sentence? (naming a term without explaining → fail) 4. **Number anchors**: does every number/angle state "from what to what"? 5. **Clean language**: mixed-in English, AI-flavored wording, awkward colloquialisms → points deducted 6. **Image-text consistency**: do the images cover all key steps of the body text? ## Scoring & Outcomes - 5 stars: a zero-background reader can retell it