iamzifei/aikey / skills/aikey-flow
aikey-flow
VerifiedAI KEY·哪里会被划走 ——口播稿「划走点」专项。替观众把稿子听一遍:每一拍观众手里有什么、在等什么,断在哪一拍 —— 等的没来、手里有了还在给、正在给的听不清。诊断完默认主动问是否标记式改稿。不负责发布红线和能不能发(→ aikey-review)、内容戳不戳人(→ aikey-resonate)、单独换开头或标题(→ aikey-hook / aikey-title)。 触发方式:/aikey-flow、/会不会划走、/顺稿、/aikey-顺稿、「这稿子顺不顺」「哪里会划走」「逻辑有没有断」「读起来卡不卡」「完播会不会掉」 Drop-off check for talking-head scripts: walk the script as a listener, beat by beat — what the viewer holds, what they are waiting for, and where that chain breaks. Not for: red lines and publish-readiness (aikey-review), whether it resonates (aikey-resonate), replacing just the opening or title (aikey-hook / aikey-title). Trigger: /aikey-flow, "does this script flow", "where will viewers drop off", "check retention risk" —— AI KEY · 不给公式,给判据。每条规则都标了实测代价。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.0 KB |
| evals/README.md | markdown | 1.1 KB |
| evals/sample-01-白板稿.md | markdown | 1.9 KB |
| evals/sample-01-答案.md | markdown | 1.4 KB |
| evals/sample-02-口播稿.md | markdown | 1012 B |
| evals/sample-02-答案.md | markdown | 1.2 KB |
| references/规则卡.md | markdown | 4.7 KB |
evals/README.md
View on GitHub# aikey-flow · 评测样本
每个样本一对文件:`sample-NN-<描述>.md`(给执行者)和 `sample-NN-答案.md`(只给评分者)。
执行者是一个**没看过答案的新 context**,拿 SKILL.md + 样本跑出报告,报告存到 `results/sample-NN-v<版本>.md`。
## 记录
| 样本 | 版本 | 命中 / 满分 | 误报 | 结论 |
|---|---|---|---|---|
| 01 白板稿 | v0.1.0 | 4 / 4 | 0 | 基线 |
| 01 白板稿 | v0.2.0 | 4 / 4 | 0 | 通过(≥ 基线)。另多报两处「等的没来」:原因三只给了一个字没给样子 · 结尾没接开头「看自己中了哪一个」的钩,均成立 |
改 SKILL.md 的行为之后,先跑一遍样本再提交;命中低于上一版就不许发。
| 01 白板稿 | v0.3.0(接规则卡) | 4 / 4 | 0 | 通过 |
| 02 口播稿 | 无技能(对照组) | 2 / 3 | 0 | 1 | 抓到 D1、D2;**D3(结尾第三次重复同一个意思)判成了金句**;另外主动建议补一句行动指令,本家族禁止 |
| 02 口播稿 | 当前版 | 3 / 3 | 0 | 对照组 2/3 → 3/3:D3「第三次重复」也抓到了,并说明为什么不判会走;没建议加行动指令 |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |