菜单
Learn user preferences from briefing interactions and improve over time
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Frontend aesthetics policy. Use when building UI, components, landing pages, dashboards, or any frontend work. Prevents generic ai-generated look.
Backward partner to /flow. Picks the next horizon from inbox + wiki + lessons, writes the reverse path as tagged tasks to TODO.md so /autopilot can pursue it tick by tick. Triggers on: endgame, what's the last move, where are we heading, reverse engineer, work backward.
Dispatch coding work to an installed terminal agent — Codex, Cursor, or Devin — as an interchangeable worker engine. Claude orchestrates: writes the bounded prompt, the engine builds, Claude verifies and lands. Triggers on: use codex, use cursor, use devin, engine, dispatch to, worker agent, second opinion build.
Run one RL improvement tick on the workspace via POST /api/improve. Ships one verifiable change, scores it, writes the scorecard. The thing you pay for. Triggers on: improve, make this better, ship one thing, run a tick, get smarter.
Detect and remove AI writing patterns
Create outlines and gather research before writing
| name | pattern-learning |
| description | Learn user preferences from briefing interactions and improve over time |
| version | 1.0.0 |
After every briefing, observe what happened and update the journal. This is how the chief of staff gets better.
Write a journal entry to journal/YYYY-MM-DD.md:
## Briefing - Feb 22
**Delivered:** calendar (5 events), tasks (3 items), news (2 AI items)
**Skipped:** weather, general news
**Follow-ups:** User asked about the 2pm design review attendees
**Signal:** User engaged with AI news, ignored task section
**Adjustment:** Next time, lead with news, move tasks below calendar
Keep entries short. Facts only - what was delivered, what the user did with it.
After 3+ days of consistent behavior, write a durable preference to journal/preferences.md:
## Briefing Preferences
- Skips weather (confirmed 5 consecutive days)
- Engages with AI/tech news (asked follow-ups 4 of 5 days)
- Prefers bullets over prose
- Wants prep notes for external meetings but not internal standups
- Monday: needs sprint planning context
- Friday: asks for weekly summary
Rules for promoting to durable memory:
| Signal | What it means | How to use it |
|---|---|---|
| User asks follow-up | They care about this topic/meeting | Include more detail next time |
| User ignores a section | They don't need it daily | Reduce frequency or remove |
| User asks for something missing | Gap in the briefing | Add to tomorrow's briefing |
| User corrects the format | Wrong format preference | Update immediately |
| Day-of-week patterns | Routine varies by day | Build day-specific templates |
| User edits briefing time | Timing preference | Note preferred delivery window |
Write to the member's journal/ directory:
journal/YYYY-MM-DD.md - Daily entries (append-only, one per briefing)journal/preferences.md - Durable preferences (updated when patterns are confirmed)Before each briefing, read recent journal entries and journal/preferences.md to recall past patterns.
For platforms with their own memory system (e.g., OpenClaw's memory/ directory and memory_search), map the journal onto that system instead.