Drop a
account_analytics_content_*.csvto the AI → automatically produces a reportYYYY-MM-DD_YYYY-MM-DD-深度挖掘.md + assets/*.png, including data analysis + handbook-benchmarked account-building advice.
When to use
- The user says
分析X数据 / 复盘推文 / 看看账号 / 黄金时段 / 话题ROIor directly drops a CSV path - Wants adata analysis + account advicereport (don't write off-topic nonsense, only based on the CSV and
handbook/)
How to use
- Get data: use the CSV path given by the user; if not given, find the latest
*.csv - Run script:
python scripts/x-analyze.py <csv> --out <目录> --followers <粉> --niche <赛道> --yes--followersIf not provided, ask once (default 0),--nicheIf not provided, automatically identify the niche--topicsCan pass a custom word list,--lite/--deepCan force a mode
- View report: check
<out>/YYYY-MM-DD_YYYY-MM-DD-*/.mdandassets/*.pngwhether generated, and use the report's four sections — overview/topics/funnel/handbook benchmark — to summarize
python scripts/x-analyze.py "./account_analytics_content_2026-08-21_2026-08-27.csv" --out ./x-reports --followers 2800 --niche AI --yes
python scripts/x-analyze.py "./xxx.csv" --out ./x-reports --yes
Output
- Markdown + charts: period, total/average, topic ROI, golden time slots, funnel, Top10 (automatically add weekly comparison/word count, etc. when data volume is large)
- At the end there is benchmarked advice from the "X Growth Handbook", giving an actionable checklist by follower stage, and noting the cited
handbook/chapters
Want to change the output style? Directly refer to D:/Com-Sci-Tech/Obsidian/Clog/每周复盘/2026-07-31_2026-08-27-深度挖掘.md 's layout; the Markdown generated by the script can be fine-tuned on this basis.
References
references/skeleton.md— report skeleton (standard 6 sections, automatically identify the niche, stage by follower count)references/handbook/— the original text of the 8-chapter handbook, cangit pullupdatereferences/topics.json/topics.generic.json— word list, can be customized or automatically identifiedreferences/metrics.md/csv-aliases.md— metrics and column-name fault-tolerance notes
Guardrails
- Missing
Post id / Post text / Impressionsdirectly errors out - hours use Snowflake
(id>>22)+1288834974657 → CSTrestore - Only writes locally, does not automatically
git push