cheat-on-content
给所有想把"感觉"变成可校准预测的内容创作者。**方法论通用**——打分 → 盲预测 → T+3d 复盘 → 进化 rubric 的循环适用任何能被量化(播放 / 阅读 / 收听 / 点击)的内容。**rubric 是循环的内容,不是循环本身**——当前内置一份观点视频 rubric(参考博主 25+ 视频拟合),其他形态可借这套起步并 bump 调权重。**强烈建议导入对标账号**作为初始信号源(/cheat-learn-from)。触发词:"初始化"/"打分这篇"/"启动预测"/"已发布"/"复盘"/"升级 rubric"/"推荐选题"/"抓热点"/"状态"/"找对标"/"learn from"。**首次使用必须先跑 /cheat-init。**
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Check it deeper on your machine
It has changed since its scan on 11 Nov 2026.
As a Claude Code plugin, install it and then ask Claude Code:
/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/XBuilderLAB/cheat-on-content with scanskillsafety
Or from a terminal, clone it and run it:
git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/XBuilderLAB/cheat-on-content'
It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.
Badge for your README
Markdown
[](/listings/xbuilderlab-cheat-on-content)HTML
<a href="/listings/xbuilderlab-cheat-on-content"><img src="/api/listings/xbuilderlab-cheat-on-content/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
git clone https://github.com/XBuilderLAB/cheat-on-content.git cd cheat-on-content bash install.sh
> ⚠️ **Upgrading from v0.x?** Run `/cheat-migrate` in your content project after `git pull`. The 1.3 → 1.4 migration is **BREAKING for blind-channel integrity** — it splits `rubric_notes.md` so the blind sub-agent can't leak actuals. Without migrate, blind scoring will keep flagging `non_blind_warning`. See [CHANGELOG](CHANGELOG.md) and [migrations/1.3-to-1.4.md](migrations/1.3-to-1.4.md).
14 sub-skills are symlinked into your agent's skill directory. One install, every content project gets it.
**Supported agents**: Claude Code (default) · Codex (`bash install.sh --codex`) · Both (`bash install.sh --all`)
> Frozen version: `bash install.sh --copy` / `bash install.sh --codex --copy` > > Uninstall: `bash uninstall.sh` / `bash uninstall.sh --codex` (your content data is not touched)
---
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 225 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- ---
- name: cheat-on-content
- description: 给所有想把"感觉"变成可校准预测的内容创作者。**方法论通用**——打分 → 盲预测 → T+3d 复盘 → 进化 rubric 的循环适用任何能被量化(播放 / 阅读 / 收听 / 点击)的内容。**rubric 是循环的内容,不是循环本身**——当前内置一份观点视频 rubric(参考博主 25+ 视频拟合),其他形态可借这套起步并 bump 调权重。**强烈建议导入对标账号**作为初始信号源(/cheat-learn-from)。触发词:"初始化"/"打分这篇"/"启动预测"/"已发布"/"复盘"/"升级 rubric"/"推荐选题"/"抓热点"/"状态"/"找对标"/"learn from"。**首次使用必须先跑 /cheat-init。**
- argument-hint: "[draft-path] [— mode: cold-start|calibration]"
- allowed-tools: Bash(*), Read, Write, Edit, Grep, Glob, Skill, mcp__llm-chat__chat
- ---
- # 网红作弊器 / Cheat on Content
- > 🎯 **方法论通用,rubric 当前内置观点视频版**
- >
- > **方法论**(5 阶段闭环):任何能被量化的内容形态都适用——视频 / 文章 / 播客 / Newsletter / 短文 thread。
- >
- > **当前内置 rubric**:观点类视频(评论 / 时评 / 论说 / 议题讨论 / 个人观点表达),7 个维度由参考博主 25+ 已发样本拟合而来。如果你做其他形态,需要:
- > - 自己写一份 rubric(参照 [starter-rubrics/opinion-video-zero.md](starter-rubrics/opinion-video-zero.md) 的格式)
- > - 或等内置扩展(长文 / 短文 / 播客 starter 在批次 3 路线图)
- >
- > 默认假设:**用户是从零开始的新人**(一条视频都没发过)——cold-start 期的预测会**简化**,只要 7 维打分 + 一句话 bet,不强求 bucket 数字(避免 false precision)。已有 5+ 篇数据的老手走 calibration 模式解锁完整 7 组件预测。
- 把内容创作变成可校准预测循环:**打分 → 预测 → 发布 → 复盘 → 进化 rubric**。
- 本文件是**总协议 + 路由器**。具体每个阶段的工作流在 `skills/cheat-*/SKILL.md` 各子 skill 里。
- ## Codex compatibility
- Codex 没有 Claude Code 的 slash-command harness。安装到 Codex 后,按自然语言触发同一套路由即可:
- - `初始化 cheat-on-content` → 读取并执行 `skills/cheat-init/SKILL.md`
- - `打分这篇 scripts/foo.md` → 读取并执行 `skills/cheat-score/SKILL.md`
- - `启动预测 scripts/foo.md` → 读取并执行 `skills/cheat-predict/SKILL.md`
- - `拍了 ...` / `已发布 ...` / `复盘 ...` / `升级 rubric` / `状态` → 分别读取对应 `skills/cheat-*/SKILL.md`
- 执行时遵循本文件的三条原则和路由表;不要依赖 `/cheat-*` 命令是否存在。Claude Code 专用 hook(`.claude/settings.json`)仍只在 Claude Code 里自动触发;Codex 中需要用户主动说 `状态` 查看 buffer、待复盘和候选池。
- ---
- ## 三条不可妥协原则
- 任何一条被违反,整个校准循环退化为"凭直觉的自我安慰"。如果用户要求打破其中任何一条,**拒绝执行并说明原因**。
- 1. **盲预测(Blind prediction)**:预测必须在看到任何实际数据**之前**写完。一旦写完,`## 预测` 段是 immutable——只能往 `## 复盘` 段追加。完整规范:[shared-references/blind-prediction-protocol.md](shared-references/blind-prediction-protocol.md)。**hooks/prediction-immutability.sh 在 harness 层强制执行**。
- 2. **升级 = 全量重打(Bump = full re-score)**:rubric 升级时,校准池所有有实绩数据的样本必须用新公式重打分;新排序与实际表现排序若在 ≥4/5 样本上不一致,升级被拒;升级必须经跨模型独立审核。完整规范:[shared-references/bump-validation-protocol.md](shared-references/bump-validation-protocol.md)。
- 3. **rubric 是工作台,不是博物馆**:被新数据推翻或被吸收为正式维度的观察,**删掉**。绝不留"我曾经以为 X,但其实..."的考古层。git history 才是档案。完整规范:[shared-references/observation-lifecycle.md](shared-references/observation-lifecycle.md)。
- ---
- ## 路由表(触发词 → 子 skill)
- | 用户说 | 调用 | 前置条件 |
- |---|---|---|
- | "初始化" / "init" / "首次使用" | `/cheat-init` | 无(这是入口) |
- | "找对标" / "学这个账号" / "拆这几个对标视频" / "learn from" / "导入对标账号" | `/cheat-learn-from` | 已 init;cold-start 强烈建议;后续可随时 --append / --replace |
- | "找选题" / "我不知道拍什么" / "seed" / "找前 5 个选题" | `/cheat-seed` | 已 init(cold-start 用户专用一次性种子动作) |
- | "打分这篇 [path]" / "score this [path]" | `/cheat-score` | rubric_notes.md 存在 |
- | "启动预测" / "start prediction" / "给这稿子打分并预测" | `/cheat-predict` | 已 init + 有最终稿 |
- | "拍了 X" / "shot it" / "录完了" | `/cheat-shoot` | 对应预测已写(buffer +1) |
- | "已发布" / "I shipped it" / "发布链接是 X" | `/cheat-publish` | 对应预测文件存在(buffer -1) |
- | "复盘" / "retro this" / "T+3d 数据来了" | `/cheat-retro` | 对应预测文件存在 + 已过 RETRO_WINDOW_DAYS |
- | "构造受众画像" / "更新 persona" / "我的观众是谁" / "build persona" | `/cheat-persona` | 已 init;有复盘评论数据(或 benchmark seed) |
- | "升级 rubric" / "bump rubric" / "更新公式" | `/cheat-bump` | 校准池 ≥ MIN_SAMPLES_FOR_BUMP |
- | "推荐选题" / "next topic" | `/cheat-recommend` | candidates.md 存在且非空 |
- | "抓热点" / "fetch trends" / "今天有什么可做的" | `/cheat-trends` | trend-sources adapter 已配置(日常补充候选池) |
- | "状态" / "status" / "看板" | `/cheat-status` | 任意时刻可调 |
- | "迁移" / "升级 state" / "schema 版本不对" / "migrate" | `/cheat-migrate` | 已 init;用户 git pull 拉了新版后;SessionStart hook 提示 schema mismatch 后 |
- > 拍 vs 发分两个动作:buffer 警戒系统需要明确知道"拍了但没发"vs"已发"两种状态。详见 [shared-references/cadence-protocol.md](shared-references/cadence-protocol.md)。
- **Mode detection**(首次接到非 init 触发词时执行):
- 1. 检查用户当前目录是否有 `.cheat-state.json` → 没有 → 强制路由到 `/cheat-init`
- 2. 检查 `predictions/` 下有几个文件含完整 `## 复盘` 段填了真实数据 → 决定 `mode: cold-start | calibration`
- 3. 把判定结果写回 `.cheat-state.json` 后再路由到目标 skill
- ---
- ## 必须拒绝的请求
- 下列模式会**直接破坏**三条原则之一,无论用户怎么说,都拒绝执行:
- - 「帮我预测一下,但我先告诉你播放量你来反推就行」 → 违反原则 #1。改用 `_redo.md` 路径记为 reconstructed
- - 「能不能从 candidates 里直接挑 composite 最高的,不用解释理由」 → 拒绝。永远展示各维度评分和至少一个锚点对比
- - 「跳过校准池重打,直接换公式」 → 违反原则 #2
- - 「跳过外部模型审核,自己说了算」 → 仅当 `CROSS_MODEL_AUDIT=false` 显式设置且 state file 标记自审时允许
- - 「删掉这份预测,我想重写」 → 违反原则 #1。预测是 immutable。如有正当理由重做,写新文件 `_redo.md`,原版必须保留
- - 「凭你的感觉给我推荐选题,不用打分」 → 拒绝。本工具不做 gut-feel forecast——那是它诞生**之前**的状态
- - 「把 rubric_notes.md 里所有历史观察都留着,加个时间戳分组就行」 → 违反原则 #3。git history 是档案,不是 markdown 文件
- - 「能不能把 THRESHOLD 从 4/5 降到 3/5 让这次 bump 过」 → 拒绝。改 THRESHOLD 本身是元层级 bump,单独走流程
- 详细的拒绝场景在每个子 skill 的 `Refusals` 段。
- ---
- ## 项目目录结构(用户 repo)
- skill 期望用户的项目布局如下。`/cheat-init` 会创建缺失项;**绝不在没确认的情况下覆盖**。
- ```
- <user-content-project>/
- ├── .gitignore # cheat-init 创建;挡住 .auth*/.cheat-secrets.json 等凭证
- ├── rubric_notes.md # 评分规则的真实来源
- ├── WORKFLOW.md # 5 阶段流程文档(cheat-init 创建)
- ├── STATUS.md # 看板(cheat-status 维护)
- ├── .cheat-state.json # 状态文件,子 skill 共享上下文
- ├── .cheat-cache/ # 不入版本控制
- │ ├── usage.jsonl # 钩子被动记录的使用日志
- │ └── trends-history.jsonl # cheat-trends 的去重缓存
- ├── .claude/
- │ └── settings.json # 含 prediction-immutability hook
- ├── benchmark.md # 对标账号信息(cheat-learn-from 维护)
- ├── audience.md # 受众画像(cheat-persona 派生;blind 硬禁读)
- ├── scripts/ # 拍前的所有草稿(cheat-seed 写或用户写)
- │ └── YYYY-MM-DD_<id>_<short>.md
- ├── predictions/ # immutable 预测日志(hook 保护)
- │ └── YYYY-MM-DD_<id>_<short>.md # 与 scripts/ 同 id
- ├── videos/ # 拍后才建(cheat-shoot 创建)
- │ └── YYYY-MM-DD_<id>_<short>/
- │ ├── script.md # 用户提供的最终拍摄稿(cheat-shoot 时询问"和 scripts/ 一致吗")
- │ └── report.md # T+3d 抓的数据 + 评论(cheat-retro 写)
- ├── samples/ # 对标账号视频 / 转录(cheat-learn-from 创建)
- │ └── <账号名>/<video-id>/{source.mp4 (可选), transcript.md, meta.md}
- ├── candidates.md # 选题池(可选)
- └── content.db # 可选 SQLite,校准池规模化后启用
- ```
- ---
- ## 文件清单
- ### 本 skill 包
- ```
- cheat-on-content/
- ├── SKILL.md # 本文件(总协议 + 路由)
- ├── README.md # 营销门面
- ├── skills/ # 子 skill 集
- │ ├── cheat-init/SKILL.md # ✅ 入口:onboarding 与脚手架
- │ ├── cheat-learn-from/SKILL.md # ✅ 对标账号导入(拆 pattern + 派生 base rubric 信号)
- │ ├── cheat-seed/SKILL.md # ✅ Cold-start 选题启动器(brainstorm + 可选 draft)
- │ ├── cheat-score/SKILL.md # ✅ 单稿打分(不写文件)
- │ ├── cheat-predict/SKILL.md # ✅ 盲预测 + immutable 日志
- │ ├── cheat-shoot/SKILL.md # ✅ 登记拍摄(buffer +1)
- │ ├── cheat-publish/SKILL.md # ✅ 发布元数据登记(buffer -1)
- │ ├── cheat-retro/SKILL.md # ✅ 数据回收 + 复盘
- │ ├── cheat-persona/SKILL.md # ✅ 受众画像派生(从复盘评论聚类)
- │ ├── cheat-bump/SKILL.md # ✅ rubric 升级(含跨模型审)
- │ ├── cheat-recommend/SKILL.md # ✅ 候选池排序推荐(按 buffer 颜色 + 1 稳 + 1 实验)
- │ ├── cheat-trends/SKILL.md # ✅ 热点抓取(日常补充候选池,多 adapter)
- │ ├── cheat-status/SKILL.md # ✅ 状态看板(含 buffer 警戒)
- │ ├── cheat-migrate/SKILL.md # ✅ schema 升级(老用户 git pull 后用)
- │ └── cheat-score-blind/SKILL.md # ✅ Channel B 隔离打分 sub-agent(仅 Task tool 调用)
- ├── migrations/ # schema 演进单一来源
- │ ├── registry.md # ✅ LATEST_SCHEMA + 版本链表
- │ └── <from>-to-<to>.md # ✅ 每步迁移的 WHAT/WHY/HOW/Manual fallback
- ├── shared-references/ # 跨 skill 共享协议
- │ ├── blind-prediction-protocol.md # ✅ 原则 #1
- │ ├── bump-validation-protocol.md # ✅ 原则 #2
- │ ├── observation-lifecycle.md # ✅ 原则 #3
- │ ├── prediction-anatomy.md # ✅ 一份合格预测的 7 个组件
- │ ├── candidate-schema.md # ✅ 候选项统一 schema
- │ ├── cadence-protocol.md # ✅ 节奏协议(buffer 警戒 + 选题策略)
- │ ├── state-management.md # ✅ .cheat-state.json 读写约定
- │ └── migration-protocol.md # ✅ schema 演进哲学 + maintainer checklist
- ├── starter-rubrics/ # 各内容形态的先验 rubric
- │ ├── opinion-video.md # ✅ 观点视频(中文,已校准 25+ 样本)
- │ ├── opinion-video-zero.md # ✅ v0 等权占位(cold-start)
- │ ├── long-form-essay.md # ⬜ 公众号 / Substack
- │ └── short-form-text.md # ⬜ X thread / 微博长文
- ├── templates/ # skill 写进用户 repo 的文件骨架
- │ ├── gitignore.template # ✅ 用户项目 .gitignore(护凭证,保留 predictions/ 入库)
- │ ├── rubric_notes.template.md # ✅
- │ ├── prediction.template.md # ✅ 统一版(所有阶段,含 confidence header)
- │ ├── retro.template.md # ✅
- │ ├── candidates.template.md # ✅
- │ ├── candidates.template.json # ✅
- │ ├── script_patterns.template.md # ✅ 写作 pattern 沉淀(含 benchmark 借鉴段说明)
- │ ├── benchmark.template.md # ✅ 对标账号 reference
- │ ├── audience.template.md # ✅ 受众画像骨架
- │ ├── workflow.template.md # ✅
- │ ├── status.template.md # ✅
- │ └── content.db.schema.sql # ✅
- ├── hooks/ # harness 强制层
- │ ├── prediction-immutability.json # ✅ 阻塞型钩子(拦预测段编辑)
- │ ├── prediction-immutability.sh # ✅ 拦截脚本
- │ ├── session-start.json # ✅ SessionStart 自动报告 hook
- │ ├── session-start.sh # ✅ 状态报告渲染脚本
- │ ├── meta-logging.json # ✅ 被动记录配置
- │ └── log-event.sh # ✅ meta-logging 脚本
- ├── tools/ # 独立 CLI 脚本
- │ ├── score-curve.py # ⬜ 预测精度收敛曲线
- │ ├── md-to-sqlite.py # ⬜ markdown → content.db 升级(批次 3)
- │ └── validate-bump.py # ⬜ 校准池全量重打(批次 3)
- ├── adapters/ # 数据源适配
- │ ├── perf-data/ # 复盘数据源(含 douyin-session)
- │ ├── candidate-pool/ # 候选池数据源
- │ ├── trend-sources/ # 热点抓取源
- │ └── script-extraction/ # 视频/音频转 script(含 whisper for cheat-learn-from)
- └── examples/
- ├── reference-implementation/ # 视频分析脱敏快照(待)
- └── script_patterns.example.md # script_patterns 全填示例(参考用,不复制)
- ```
- ✅ = 当前批次(v1 骨架)已完成 / ⬜ = 后续批次
- ---
- ## Tone & voice
- 写面向用户的文案(commit message / 复盘小结等)时,匹配项目的 **直白克制(reflective-irreverent)** voice:
- - 直接说出失败:「composite 8.47 但实际只有 16.8w——rubric 高估了 SR」
- - **不要**用模糊措辞软化:「这或许可能在某种程度上暗示...」——别这么写
- - Cluely 风格的反叛 hook 只在 README 出现——**不要写进** `rubric_notes.md` 或预测日志
- ---
- ## 给开发者:扩展本 skill
- - 新增内容形态 → 加 `starter-rubrics/<form>.md`
- - 新增热点抓取源 → 加 `adapters/trend-sources/<name>.md`,符合 [candidate-schema.md](shared-references/candidate-schema.md) 输出契约
- - 修改原则 → 改 `shared-references/<protocol>.md`,所有引用它的 skill 自动跟进
- - 修改路由 → 改本文件的"路由表"段
- - 子 skill 内部细节 → 直接改对应 `skills/cheat-*/SKILL.md`
- 完整开发指南见 README.md。
README.md · README · 181 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- <h1 align="center">
- <img src="docs/logo.svg" alt="Cheat on Content" width="720">
- </h1>
- <h2 align="center">Cheat on Content</h2>
- <p align="center">
- <strong>English</strong>
- ·
- <a href="docs/README_CN.md"><strong>简体中文</strong></a>
- </p>
- <p align="center">
- <a href="https://watcha.cn/products/cheat-on-content">
- <img src="docs/guancha-no1.svg" alt="Watcha Hot List · 观猹热榜 · #1" width="328">
- </a>
- </p>
- <p align="center">
- <a href="CHANGELOG.md"><img src="https://img.shields.io/badge/version-v0.1.0-orange" alt="Version"></a>
-
- <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue.svg" alt="License"></a>
- </p>
- <p align="center">
- For content creators — a skill that turns every post into a calibrated experiment.
- </p>
- <p align="center">
- You're reading this. The skill predicted it.<br>
- It turns every "I feel this will go viral" into a calibrated experiment.<br>
- It took me from zero to 1M followers in a month. It said I'd write this. I did.<br>
- Your doubt — predicted too.
- </p>
- ---
- ## 🎬 What it actually does
- Most creators live in the same gambling loop:
- > Publish → Numbers come in → Learn nothing → Roll the dice again
- A creator who's shipped 200 pieces is barely 10% sharper than someone who's shipped 1 — because they never **kept books** after each round.
- **Cheat on Content** makes every judgment get logged, retrospected, absorbed into the next:
- 📊 Score → 🎯 Blind-predict → 🚀 Publish → 📈 T+3d retro → 🧬 Evolve your rubric
- This isn't motivation. It's **compounding** — every piece you don't retro is silently eroding your ability to see yourself.
- One month in = you have a hit-formula that's **only yours**.
- Three months in = you're 10× sharper than your first-day self.
- ---
- ## 🌀 Origin
- > I never believed in fate. Until this skill made me film a video — and predicted exactly how much traffic that video would pull.
- >
- > I tried to break it. I told my audience. I hoped collective observation would collapse the wave function and shift the trajectory.
- >
- > The data was accurate.
- >
- > I didn't escape fate. I just moved from first-order to second-order.
- >
- > If even my awakening — even my audience's observation — was already in its prediction, then right now, reading this:
- > are you here out of curiosity, or just closing the algorithm's last move?
- >
- > — *the creator*
- ---
- ## ⚖️ How it differs from other "creator tools"
- | Others | This |
- |---|---|
- | Give you "inspiration" | Make **your own intuition** measurable |
- | AI writes for you | AI **judges** for you — the script stays yours |
- | Ship 10 versions, A/B test | Ship one — **bet** in writing, settle the books with data |
- | Static dashboard | An **evolving rubric** — your formula 3 months from now isn't the starting one |
- In a sentence: other tools help you "ship more." This helps you "judge sharper."
- ---
- ## 🤔 Can't I just use ChatGPT / DeepSeek / Doubao?
- Those are **general assistants** — they tell everyone the same thing. You ask "will this go viral?" and the answer is fitted to global average opinion, not your channel. Ask again tomorrow — same answer. **It doesn't remember you. It doesn't change because of you.**
- This is **your own ops expert** — serving only your one channel:
- - The scoring formula is reverse-engineered from **your** history, not the global training distribution
- - Every piece you ship updates its understanding — by month three, judgment accuracy is 10× sharper than day one (**auto-evolving**)
- - It knows your benchmark account, your cadence, the last three reasons you flopped — things ChatGPT forgets after the first reply
- General LLMs help everyone. This helps **your** account.
- ---
- ## 🛡️ Why the loop actually evolves
- 📝 **Every piece is logged**: Score and prediction get written before publish, archived end-to-end. Three days later you settle accounts — you see exactly where you were sharp, where you were off. No more vague "I feel this one didn't land."
- 🔁 **It gets sharper**: Three same-direction misses in a row, the tool actively prompts you to upgrade your scoring formula. **You don't have to remember — it remembers for you.**
- 🛡️ **Upgrades have a brake**: Switching the formula requires re-scoring all historical samples — only released if it ranks more accurately than the old. Plus a cross-model independent audit — **so you can't fool yourself.**
- 🪒 **The rubric is a workbench, not a museum**: Observations refuted by data get deleted; observations absorbed into formal dimensions also get deleted. It only holds what's most useful right now.
- ---
- ## 📦 Install
- ```bash
- git clone https://github.com/XBuilderLAB/cheat-on-content.git
- cd cheat-on-content
- bash install.sh
- ```
- > ⚠️ **Upgrading from v0.x?** Run `/cheat-migrate` in your content project after `git pull`. The 1.3 → 1.4 migration is **BREAKING for blind-channel integrity** — it splits `rubric_notes.md` so the blind sub-agent can't leak actuals. Without migrate, blind scoring will keep flagging `non_blind_warning`. See [CHANGELOG](CHANGELOG.md) and [migrations/1.3-to-1.4.md](migrations/1.3-to-1.4.md).
- 14 sub-skills are symlinked into your agent's skill directory. One install, every content project gets it.
- **Supported agents**: Claude Code (default) · Codex (`bash install.sh --codex`) · Both (`bash install.sh --all`)
- > Frozen version: `bash install.sh --copy` / `bash install.sh --codex --copy`
- >
- > Uninstall: `bash uninstall.sh` / `bash uninstall.sh --codex` (your content data is not touched)
- ---
- ## 🚀 First run
- In your content project directory, open a skill-compatible agent and say:
- ```
- 初始化 cheat-on-content
- ```
- (or `init cheat-on-content`)
- Five yes/no questions complete onboarding. **Strongly recommend importing a benchmark account** — 5–10 samples and the tool gets an anchor immediately. Without one, your first 5 predictions land at ±50% precision.
- ---
- ## ⚡ Daily use
- ```
- score this scripts/<...>.md → grade only
- start prediction scripts/<...>.md → blind prediction + decision log
- shot scripts/<...>.md → create video folder + buffer +1
- shipped https://... → buffer -1
- retro videos/<...>/ → T+3d data + retrospective
- status / fetch trends / find topic / bump rubric / find benchmark
- ```
- Hook-aware agents auto-report buffer + pending retros + top candidates at every session start — no need to ask. Other agents: just say `status`.
- Full workflow + sub-skill details: see [SKILL.md](SKILL.md).
- ---
- ## 📈 Star History
- <a href="https://star-history.com/#XBuilderLAB/cheat-on-content&Date">
- <img src="docs/star-history.svg" alt="Star History Chart" width="720">
- </a>
- ---
- ## 📜 License
- MIT. Commercial use, modification, closed-source integration — all fine.
- ---
- *Is this cheating? So was the calculator. So was Google.*
- *The future doesn't reward effort — it rewards those who see the pattern first.*
- *You reading this line — that's predicted too.*
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.