x-list-digest

分类: 调研与分析 | 上传者: WEI-88-WEI | 下载: 0 | 版本: v1.0（最新）

Fetch, deduplicate, filter, summarize, and archive tweets from a fixed set of X/Twitter Lists for Obsidian-based research. Use when the user asks things like “帮我获取列表推文”, wants list updates since the last fetch, wants summaries from the configured lists (星, 看, maomao, meme, 生态, 项目, 其他), or wants time-window list digests pushed to GitHub.

更新日志: Source: GitHub https://github.com/WEI-88-WEI/x-info

目录结构

当前层级: skills/x-list-digest/

📁 data/
- 📁 2026-03-13/
  - 📄 00:14:47~15:26:28.md 8.4 KB
  - 📄 15:51:32~16:42:35.md 4.9 KB
  - 📄 21:28:49~23:10:37.md 3.3 KB
- 📁 2026-03-14/
  - 📄 00:10:00~23:36:35.md 9.3 KB
- 📁 2026-03-15/
  - 📄 00:00:45~08:35:04.md 6.0 KB
  - 📄 09:53:14~23:30:54.md 7.6 KB
- 📁 2026-03-16/
  - 📄 00:18:53~09:19:37.md 6.9 KB
  - 📄 11:40:51~23:54:31.md 7.5 KB
- 📁 2026-03-17/
  - 📄 00:11:41~09:46:55.md 6.2 KB
  - 📄 10:00:26~10:29:48.md 4.5 KB
  - 📄 10:36:37~10:36:37.md 877 B
  - 📄 10:44:54~13:23:27.md 8.2 KB
  - 📄 15:19:47~21:13:58.md 9.0 KB
  - 📄 21:25:31~23:19:55.md 8.7 KB
- 📁 2026-03-18/
  - 📄 00:10:22~10:06:42.md 9.4 KB
- 📄 state.json 1.3 KB
📁 references/
- 📄 lists.md 698 B
- 📄 obsidian-format.md 2.0 KB
📁 scripts/
- 📄 build_digest.py 27.4 KB
- 📄 fetch_list.py 5.5 KB
📄 SKILL.md 4.7 KB

SKILL.md

---
name: x-list-digest
description: Fetch, deduplicate, filter, summarize, and archive tweets from a fixed set of X/Twitter Lists for Obsidian-based research. Use when the user asks things like “帮我获取列表推文”, wants list updates since the last fetch, wants summaries from the configured lists (星, 看, maomao, meme, 生态, 项目, 其他), or wants time-window list digests pushed to GitHub.
---

Use this skill to maintain a repeatable pipeline for the configured X lists.

## Workflow

1. Fetch tweets from the requested alias, or all aliases if the user does not specify one.
2. If this is the first run for an alias or the user wants a rebuild, use bootstrap and fetch the latest 30 tweets.
3. Otherwise, fetch recent tweets and keep only items newer than the alias checkpoint in `data/state.json`.
4. Deduplicate by `tweet_id` globally before writing output.
5. Filter aggressively. Keep only tweets that match the fixed tags and have clear signal quality.
6. Prioritize airdrop info, trading setups, macro signals, and notable project updates for this user.
7. Write files using Beijing time with date folders and time-window filenames: `data/<YYYY-MM-DD>/<HH:MM:SS~HH:MM:SS>.md`.
8. Inside each note, use sections for aliases that produced useful content: 星, 看, maomao, meme, 生态, 项目, 其他.
9. Keep the note lean. Do not add `总览`、模板化 `摘要`、单独 `标签` 区块或 `重点来源` 区块。
10. Add a short `全部列表总结` near the top. Write it as an overall digest of the whole batch, not as a showcase of a few picked tweets.
11. Do not use `先看：` or any similar picked-items intro.
12. Summarize the batch-level picture with concrete content from this batch: actual price moves, policy events, protocol updates, funding flows, trading setups, numbers, and conditions.
13. `全部列表总结` should be thematic synthesis, not stitched-together tweet excerpts. Rewrite and compress across multiple tweets into 2-3 concise paragraphs.
14. Prefer a stable structure when material exists: `交易主线...` / `宏观和结构性讨论...` / `机会更具体地落在...`.
15. Do not let the same tweet or the same fact pattern appear twice in `全部列表总结`, even if it matches multiple tags.
16. Do not start with bookkeeping or meta-stat lines like `这批内容主要集中在…` or `本次共整理…`.
17. Avoid vague analyst filler like `风险偏好回升` / `值得继续跟` / `情绪改善` unless tied to specific events or numbers from the batch.
18. Do not include alias-distribution lines like `空投/积分主要在…` / `交易主要在…` / `宏观主要在…`.
11. Under each alias, keep only the strongest subset after ranking. Prefer fewer, sharper items over completeness.
12. For each kept tweet, write one compact Chinese summary line. Then put that tweet's tags, posting user display name plus handle, and original link directly under the item.
13. Do not emit `判断` or `动作` lines, and do not append interpretive tails like `这条偏政策和监管信号` or `适合放进宏观环境变量里看`.
14. For non-Chinese tweets, rewrite the main content into concise Chinese instead of dumping the English original.
15. If a rewritten summary still leaves obvious English sentence fragments, drop that item rather than emitting mixed-language output.
16. If the source says `bought another 30,000 ETH via FalconX 8 hours ago`, the summary should read like `8 小时前又通过 FalconX 买入 30,000 枚 ETH` rather than leaving the English sentence intact.
17. Commit and push to the GitHub repo after generation when the user asked to save results.

## Output rules

- Always write Markdown in Obsidian-friendly format.
- Date must be a folder.
- File name must be the time window.
- All displayed dates and times must use Beijing time.
- Group content by alias inside the note.
- For alias `星`, allow more detail because it has the highest weight.
- Filter out low-value and off-topic tweets.
- Keep only tweets relevant to these tags when possible: `#airdrop` `#macro` `#trading` `#defi` `#ai` `#btc`.
- Put tags, posting user display name plus handle, and original link directly under each kept item.
- Avoid analyst-sounding filler that could apply to any tweet.
- Prefer concrete facts, numbers, timing, conditions, and changes.

## Commands

Build time-window digests:

```bash
python3 {baseDir}/scripts/build_digest.py --all
```

Bootstrap and rebuild from the latest 30 tweets:

```bash
python3 {baseDir}/scripts/build_digest.py --all --bootstrap
```

Build time-window digests and push to GitHub:

```bash
python3 {baseDir}/scripts/build_digest.py --all --commit
```

## GitHub

Use `build_digest.py --commit` to commit and push only generated digest files and the state file.

Do not push unrelated workspace files.

登录后下载/点赞/收藏 ❤ 3 | ★ 0

x-list-digest

目录结构

SKILL.md

举报内容

提示