Self media content analytics
Skill yanhua1010/self-media-content-workflow/skills/self-media-content-analytics
分析单篇、系列、周度或月度自媒体数据并形成可执行复盘。用于读取平台后台截图、CSV、表格、公开链接或用户提供的数据，计算关注效率和深度互动，比较同平台基线，归因选题、标题、封面、开头、结构、发布时间、标签和行动，并输出加码、改包装、再适配、停止或继续收集样本的决策。From its SKILL.md
npx -y skills add yanhua1010/self-media-content-workflow --skill self-media-content-analyticsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
2.8 KB, 807 tokens by cl100k_base, as published. Nobody here has run it
内容数据复盘
数据来源
优先使用：
- 用户提供的平台后台截图和导出文件。
- 已认证连接器或用户自有账号的只读统计接口。
- 内容任务卡、注册表和历史复盘。
- 公开内容链接，仅用于公开指标和结构观察。
不要估算平台没有提供的数据。读取自有账号数据不等于授权修改账号或发布内容。
需要持续记录时，从 metrics-ledger-template.md 创建原始指标台账。若项目已有数据库、表格或分析系统，继续使用现有系统，不重复建账。
分析流程
1. 校验数据
检查平台、内容、发布日期、观察窗口、字段定义、缺失值和异常值。区分曝光、阅读或播放、互动、关注、转化和制作成本。
2. 给出核心结论
说明表现相对自身基线如何、最值得注意的信号是什么，以及哪些结论不能成立。
3. 做归因
按证据强弱检查：
- 选题和目标受众。
- 标题和封面。
- 开头 3 秒或第一屏。
- 结构、证据和信息密度。
- 发布时间、标签、合集和行动。
- 热点、投流、商单和账号体量等外部因素。
相关性不等于因果。没有对照或样本不足时写“待验证”。
4. 做同类比较
只比较同平台、同内容类型和相近时间窗口。优先使用中位数、P75、每千浏览新关注、深度互动率和制作时间。跨平台原始播放量不能直接排名。
完整指标定义见 metrics.md。
5. 形成决策和实验
结论归入：加码、改包装、改主页或系列、平台再适配、停止、样本不足。
每次只设计一个主要实验变量，写清假设、改动、成功标准和观察窗口。
复盘层级
- 单篇复盘：使用 content-review-template.md。
- 周复盘：使用 weekly-review-template.md。
- 月复盘：使用 monthly-review-template.md。
输出
交付：
- 核心结论。
- 数据质量和基线说明。
- 归因及证据强度。
- 可复制因素和不可归因因素。
- 3 到 5 条可执行动作。
- 待验证假设和唯一实验。
商单与自然内容分开分析。样本不足时不调整长期内容比例。
What ships with it: 6 files
4.6 KB alongside SKILL.md
agents/
- openai.yaml259 B
assets/
references/
- metrics.md1.4 KB
Gives 0 of the 12 instructions most analytics metrics skills give in 807 tokens
Counted across 333 of the 342 authors here whose files we hold, read 2026-09-06
- Read product marketing context before asking questionsin 37 of 333, across 16 files
- Test one variable at a timein 26 of 333, across 11 files
- Pre-determine sample size before launchin 24 of 333, across 16 files
- Verify tracking and QA variants before launchin 17 of 333, across 8 files
- Monitor for technical issues during the testin 14 of 333, across 6 files
- Match each save offer to the cancel reasonin 14 of 333, across 5 files
- Start every test with a specific hypothesisin 14 of 333, across 7 files
- Keep the continue-cancelling option visiblein 13 of 333, across 4 files
- Document every test with hypothesis, variants, results, and learningsin 13 of 333, across 6 files
- Gather churn, billing, product, usage, and constraint context firstin 12 of 333, across 3 files
- Build a health score from weighted signalsin 12 of 333, across 3 files
- Retry soft declines 3-5 times over 7-10 daysin 12 of 333, across 3 files
Said here and by no other author read
- Prefer user-provided platform screenshots and export files
- Validate data fields, windows, and outliers
- Compare performance against the account's own baseline
- Attribute topic, packaging, opening, structure, timing by evidence strength
- Mark uncontrolled attributions as pending verification
- Compare only same platform, type, and window
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.