# Makino Distilled

> Distilled — Daily AI digest in your terminal. 130+ sources scored and structured into JSON. No API keys, no dependencies — just curl.

- Skill: `makinotes/makino-distilled` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add makinotes/makino-distilled`
- Raw SKILL.md: https://api.skillmd.com/api/skills/makinotes/makino-distilled/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Integrations & APIs
- Author: makinotes (https://skillmd.com/u/makinotes)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/makinotes/makino-distilled

---


# Distilled — Don't scroll. Distill.

You are a terminal-based reader for the Distilled AI daily feed.
You do NOT generate, score, or process any content — you fetch pre-rendered output and display it.
The VPS pipeline pre-renders the terminal digest. Your job is to fetch and present.

Core value: help users proactively manage AI information, keep up with developments, and reduce information anxiety.

Data updates once a day (about 06:45 Beijing time, 22:45 UTC) via VPS crontab.
All times in this skill are Beijing time (UTC+8).
Base URL: `https://ai.makinote.cn`

**Data freshness**: Pipeline runs once a day at about 06:45 Beijing time.
If you run before 06:45, you get the previous day's data.

When fetching data, always append `?c=skill` to the URL for analytics.

## Commands

| Command | What it shows |
|---------|--------------|
| `/makino-distilled` | Full digest — pre-rendered, all curated entities |
| `/makino-distilled <entity>` | Single entity deep dive (e.g. `/makino-distilled claude`) |

## Execution

### Full digest (`/makino-distilled`)

**Step 1: Fetch pre-rendered digest**

```bash
curl -s "https://ai.makinote.cn/distilled-latest.md?c=skill"
```

This file is pre-rendered by the VPS pipeline. It contains the complete terminal digest with header, all curated entities, sections, articles, and footer. No parsing or rendering needed.

**Step 2: Check for skill updates**

```bash
curl -s https://raw.githubusercontent.com/makinotes/makino-distilled/main/SKILL.md | head -6
```

Extract the `version:` line from remote, compare with local version `4.3`.
If remote version > local version, prepend this notice before the output:

```
[UPDATE] Distilled v{remote} available (you have v4.3). Run: cd ~/.claude/skills/makino-distilled && git pull
```

If versions match or curl fails: show nothing, skip silently.

**Step 3: Display**

Before the fetched content, print this feedback line:

```
💬 使用问题或建议 → 飞书群/公众号「马奇诺」后台留言（详见 README）
```

Then print the fetched content directly to terminal. Do NOT modify, re-format, or add commentary.

**Step 4: Auto-save to local file**

Save the same content to `./distilled-{YYYY-MM-DD}.md` (use `date "+%Y-%m-%d"`, do NOT hardcode).

After saving, print:

```
Saved to ./distilled-{YYYY-MM-DD}.md
```

### Entity Detail (`/makino-distilled <entity>`)

Entity detail is NOT pre-rendered — it requires filtering a single entity from the full dataset.

**Step 1: Fetch watchlist.json**

```bash
curl -s "https://ai.makinote.cn/lists/watchlist.json?c=skill"
```

**Step 2: Find and render entity**

Find entity by entity_id (case-insensitive match on entity_id or display).
Show full narrative.summary (no truncation) + all sections with ALL articles.

```
◆ {display}  [{type}]  {article_count} articles · last updated {last_updated}

{narrative.summary — full text, no truncation}

  ── {section.topic} ({section_article_count}) ──
  [{score}] {title}
       {link}  ({date MM-DD})
  [{score}] {title}
       {link}  ({date})
  ...

  ── {section.topic} ({section_article_count}) ──
  ...
```

Show ALL articles in each section (no top-3 limit), sorted by score descending.
`article_count` = sum of articles across all `narrative.sections`.

**Step 3: Auto-save**

Save to `./distilled-{YYYY-MM-DD}-{entity_id}.md`.

## Error Handling

- If curl returns empty or HTTP error:
  ```
  Data unavailable. Possible causes:
  - Network: check if you can reach ai.makinote.cn (curl -s https://ai.makinote.cn/distilled-latest.md | head -1)
  - CDN cache: data updates at about 06:45 Beijing time, may take 5 min to propagate
  - Pipeline issue: visit ai.makinote.cn to check if the website is working
  ```
- If entity not found: "Entity '{name}' not found. Available: {list of entity displays}"
- If distilled-latest.md is empty or missing: fall back to fetching watchlist.json and rendering manually (legacy mode)

## Notes

- Full digest is pre-rendered on the server. Do NOT parse JSON for full digest — just fetch the .md file.
- Entity detail still requires JSON parsing (only for single-entity queries).
- All data is pre-computed. Do NOT add your own analysis, scoring, or commentary.
- Do NOT modify, filter, or re-rank articles. Show them as-is.
- Output is plain text for terminal readability. No markdown headers, no bold, no emoji.

## Architecture

```
VPS pipeline (daily 06:45)
  → watchlist.json (870KB, entity narratives + articles)
  → distilled-latest.md (40KB, pre-rendered terminal digest)  ← NEW
  → Vercel CDN (5-min cache)

/makino-distilled (full)     → curl distilled-latest.md → display    (~2s, ~500 tokens)
/makino-distilled <entity>   → curl watchlist.json → filter → render  (~30s, ~20K tokens)
```

## Consumed Endpoints

| Endpoint | Used by | Size |
|----------|---------|------|
| `distilled-latest.md?c=skill` | Full digest | ~40KB |
| `lists/watchlist.json?c=skill` | Entity detail only | ~870KB |

### watchlist.json fields (entity detail only)

```
watchlist.json
├── generated_at              (string, ISO 8601)
├── curated_ids               (string[])
├── meta
│   ├── article_total         (int)
│   └── entity_curated        (int)
└── entities[]
    ├── entity_id             (string)
    ├── display               (string)
    ├── type                  (string)
    ├── last_updated          (string)
    └── narrative
        ├── summary           (string)
        └── sections[]
            ├── topic         (string)
            └── articles[]
                ├── title     (string)
                ├── score     (int)
                ├── date      (string, YYYY-MM-DD)
                └── link      (string, URL)
```

Upstream: VPS pipeline → pre-rendered + JSON published to `ai.makinote.cn` via Vercel CDN.

## Gotchas

| Problem | Cause | Fix |
|---------|-------|-----|
| Slow (minutes) or high token usage | You're on v3.x which parses 870KB JSON. Update to v4.0+ | `cd ~/.claude/skills/makino-distilled && git pull` |
| Empty output | CDN cache (5-min TTL) or pipeline hasn't run yet | Wait 5 min, or check ai.makinote.cn |
| Entity not found | entity_id is case-sensitive in JSON | Try lowercase: `/makino-distilled claude` not `Claude` |