Scan 待发布的内容/中短视频/, and write all videos with complete YAML data into the data stats table + archive them to the published directory.
Design Premises
- Data source: each script file's YAML frontmatter (the user has already manually filled in the platform data after publishing)
- Do not ask the user for numbers: all data is read from YAML; if a field is missing, skip that entry and tell the user to fill it in
- Batch, idempotent: can be run repeatedly; already-archived items will not be processed again
Workflow
Step 1: Scan the directory
Glob 01-内容生产/待发布的内容/中短视频/*.md, to get the list of candidate files.
No files → output 📭 没有待同步的视频 and end.
Step 2: Read and validate the YAML of each file one by one
For each file, Read its YAML and classify it according to the following rules:
Complete(all satisfied):
发布时间non-empty点赞数、收藏数、观看数、完播率all non-empty- （
涨粉数allowed to be empty)
Missing fields: publish time is empty, or any key data field is empty → add to 跳过列表, record the names of the missing fields.
Complete groupproceeds to step 3;Skip groupis left for the unified report in step 6.
Step 3: Process each complete file
For each file in the complete group, execute 3.1 → 3.5 in order:
3.1 Calculate metrics
Save/like ratio = saves ÷ likes (keep 2 decimal places)
Hit level(by number of likes):
- 5000+：✈️
- 2000-4999：🚗
- 1000-1999：🛵
- 500-999：🚲
- <500：🚶
3.2 Write to the stats table
Read 03-数据统计/数据统计表.md, use fuzzy title matching to find the corresponding row:
- Found and is
待填入→ Edit to replace with:| N | YYYY-MM-DD | 标题 | 分类 | [等级] | [点赞] | [收藏] | [收赞比] | [观看] | [涨粉] | [完播率]% | - Found but already has data → skip that row (to avoid overwriting), mark in the report
已存在数据，未覆盖 - No corresponding row found → append a new row at the end of the table
If follower growth is missing, fill in -. Completion rate is uniformly stored as 8.1% format.
3.3 Sync the topic record
Read 01-内容生产/选题管理/00-选题记录.md, find the corresponding entry.
If it is 已完成 status, update to:
- [x] [标题] - 已发布（YYYY-MM-DD）
3.4 Archive the file (with idempotency check)
Determine the category according to the classification rules in CLAUDE.md:
- AI Tool In-Depth Review: tool reviews, sponsored recommendations, feature introduction types
- AI Application Development Learning: tutorials, development experience, technical popularization types
- One-Person Company Growth Diary: personal experiences, vlogs, workflow sharing types
Target path:02-已发布内容/短视频/[分类]/[YYYY年MM月]/[名].md（YYYY年MM月 taken from the YAML's 发布时间）
Idempotency check: if a file with the same name already exists at the target path → do not overwrite, mark 已归档过，跳过, skip to 3.5.
Otherwise create the target directory and move the file.
3.5 Delete the deep-dive report
In 01-内容生产/待深化的选题/ , look for a deep-dive report with the same name; if it exists, delete it.
Step 4: Batch Report
After processing all files, output in a unified manner (do not output fragmented results file by file):
✅ 同步完成：N 条
【已归档】
1. [标题1] · [等级] 点赞X 收藏X · 收/赞比X · 观看X · 完播率X%
→ 02-已发布内容/短视频/[分类]/[年月]/
2. [标题2] · ...
【已存在跳过】
- [标题3]：02-已发布内容/... 已存在同名文件
【缺字段跳过】
- [标题4]：YAML 缺 [发布时间, 点赞数]
- [标题5]：YAML 缺 [完播率]
请补齐 YAML 后重跑本 skill。
When a group is empty, omit the corresponding subsection.