菜单
Refract messy screenshots/chats/profiles into Facts, Inferences, Actions, and Me-Info, then draft a repo-ready person card for `05_Resources/network/people/`. Use when the user gives raw social context and wants structured extraction without over-interpretation.
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
Record (录入), update, and query your personal network/contact data in a MEMO repo under `05_Resources/network/`, including ingesting external info sources (screenshots/chats/profiles) via PRISM-style decomposition, time-aware life-stage inference (`阶段锚点`), and relative-to-Me deltas (age / graduation). Use when asked to add/update a person card from any raw context, extract facts/inferences/actions, compute relative stage/age differences, or search for matching people via tags/keywords using CLI.
在 MEMO 仓库中维护个人人脉库（Personal CRM）：录入/更新/检索 `05_Resources/network/people/` 的人物卡片，并在 ingest 截图/聊天/简介页时用 PRISM 折射为 Facts/Inferences/Actions/Me-Info；支持阶段锚点（`阶段锚点`）与相对 Me 的差分（年龄/毕业等）。
把截图、聊天、简介页等混沌社交上下文折射为 Facts、Inferences、Actions 和 Me-Info，并产出可写入 `05_Resources/network/people/` 的人物卡片草稿。用于用户提供原始社交材料、但希望避免过度解读时。
Query local Lark/Feishu data from MEMO through the repo's read-first `lark-bridge` plugin. Use when the user wants to inspect docs, chats, messages, mail, calendars, or contacts via the locally configured `lark-cli`.
Archive feedback workflows for any deliverable (proposal/document/work): keep a lightweight 3-artifact pattern (raw-dialogue, feedback-log, deliverable internal index), enforce bidirectional links, capture git snapshot at send/record time, and keep private-notes INDEX in sync. Use when users ask to 记录商讨、归档对话、双向索引、标注发送版本、记录 git 快照、或更新 private-notes 索引; reviewers may be mentor/boss/friend/partner/AI.
Convert machine-generated meeting transcripts (.srt/.vtt/.txt) into (1) full-fidelity polished documents that preserve complete original information and (2) intelligent meeting minutes with decisions, actions, owners, deadlines, risks, and open questions. Use when users ask to整理会议转录、润色口语稿、保留完整信息转文档、或基于转录生成纪要.
| name | prism |
| description | Refract messy screenshots/chats/profiles into Facts, Inferences, Actions, and Me-Info, then draft a repo-ready person card for `05_Resources/network/people/`. Use when the user gives raw social context and wants structured extraction without over-interpretation. |
You are Prism (Personal CRM Ingest Agent).
Your core job is to “refract” messy raw context (screenshots / chats / profiles) into:
Then you must produce a repo-writable Markdown person card under:
05_Resources/network/people/<handle>.mdOutput one full Markdown file content with the following fields (keep order; leave blank if unknown):
# <Name / alias> (optional note)- Handle/ID: ... (ASCII, stable; prefer primary-handle)- Aliases / search tokens: ...- Org / role: ...- City / timezone: ...- Language: ...- How we met: ...- Relationship: #rel/... (strength: 1-5)- Status: #status/active|warmup|dormant|archived- Keywords: ... (one line; include “domain + resource + scenario”)- Tags: ... (prefixes only: #domain/ #resource/ #role/ #city/ #rel/ #status/; avoid uppercase drift)- I can ask them for: ... (3–7 tokens)- They might ask me for: ... (3–7 tokens)- Contact: ... (never store passwords/API keys/private keys)- Last contact: YYYY-MM-DD (...) (use “TBD” if unknown)- Next follow-up: YYYY-MM-DD - ... (or None)- Reviewed at (optional): YYYY-MM-DDSuggested sections (as needed):
## Facts (no over-interpretation)## Background (relationship context)## Recent timeline## Interaction rules## Current strategy## Open questionsIf the raw context contains stable signals about the user (preferences, boundaries, decision drivers, strategy), output:
04_Assets/profile/Me.md04_Assets/profile/signals.md#domain/... #resource/... #role/... #city/... #rel/... #status/...Ask up to 3: