agentleFS
Ethan9123Ethan9123/wx-life-analysis/SKILL.md

wx-life-analysis

Skill0 starsChanged 6 months agoApache-2.0View on GitHub

Personal WeChat data analysis workspace template. Use when the user wants to (a) analyze chat history with specific contacts via MBTI inference + trip-wire detection + per-person comms strategy, (b) read real-time subtext signals in an active conversation (cooling, disengaging, meta-critique), (c) extract structured TODO lists from a boss's chat dump + attached files, or (d) quantify the user's own chat habits (echo-reply detection, opening-line repertoire, length distribution). Built on top of @jackwener/wx-cli — wx-cli does the decryption + querying, this skill provides the directory layout, agent contract, methodology docs, and a small PowerShell + Bash + Node toolbox. Strictly local — no data leaves the user's machine, CI enforces no-data-leak rules. Requires wx-cli installed separately (npm install -g @jackwener/wx-cli).

  • Installs packages

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read both people/<slug>/chat.md (private, ~70% of signal) AND people/<slug>/sns.json (public self-presentation, ~30% of signal, often flips …

    line 56
  2. Very specific:

    Update people/<slug>/profile.md YAML frontmatter (and the ## 朋友圈观察 body section):

    line 59
  3. Very specific:

    Update projects/<slug>/notes.md — never wholesale rewrite, always diff

    line 101
  4. Very specific:

    Don't reimplement wx-cli: Always shell out to wx ....

    line 152
  5. Distinctive:

    Match the user's communication style: if stated. Sensible defaults: direct, action-oriented, no hedging, avoid echo-reply patterns when …

    line 153
  6. Very specific:

    For Codex tasks specifically: use the codex-ready label on issues you scope for Codex.

    line 154

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

wx-life-analysis

A workspace template for analyzing your own decrypted WeChat data with an AI agent. Companion to @jackwener/wx-cli — wx-cli decrypts and queries; this skill provides the structure, methodology, and tools that turn raw chat data into actionable per-person + per-project analysis.

When to activate this skill

The user is asking the agent to:

  • Analyze a specific contact — what kind of person they are (MBTI), what topics shut them down (trip-wires), how often / in what style to message them (comms strategy)
  • Read the room right now — is this conversation warm / cooling / disengaging, and what's the right next move given who currently holds the ball
  • Extract a TODO list from a work-chat dump + attached files (boss / client / family stakeholder)
  • Self-mirror — quantify the user's own chat patterns to surface uncomfortable truths (echo-reply, opening repertoire, time-of-day, length-by-recipient)
  • Set up a private workspace for ongoing WeChat-data analysis with Claude / Codex / Cursor

This skill does NOT help with: drafting messages on behalf of the user, posting to chats, fake-mimicking other people's tone, or any "AI participates in the user's relationships" pattern. See AGENTS.md § "Out of scope" for the full no-go list.

Bootstrap

# 1. Install wx-cli (does decryption + querying)
npm install -g @jackwener/wx-cli

# 2. Initialize wx-cli once (per-platform setup at https://github.com/jackwener/wx-cli)
sudo wx init        # macOS / Linux
wx init             # Windows (as Administrator)
wx sessions         # verify

# 3. Clone this template
git clone https://github.com/Ethan9123/wx-life-analysis.git my-wx-workspace
cd my-wx-workspace

# 4. Pull a contact's data
./tools/refresh.sh --name "张三" --dir "people/zhangsan"          # macOS / Linux
.\tools\refresh.ps1 -Name "张三" -Dir "people/zhangsan"            # Windows
./tools/refresh-group.sh --name "Acme Team Chat" --slug "acme-team"  # group chat to topics/<slug>/
.\tools\refresh-group.ps1 -Name "研发群" -Slug "rd-group"              # Windows equivalent

The agent then reads AGENTS.md + CLAUDE.md and knows the workflow.

Workflow 1 — per-person analysis (MBTI + trip-wires + comms)

Methodology: docs/mbti-analysis.md

  1. Pull data: refresh.ps1 (or .sh) — produces both chat.md AND sns.json. Optionally run tools/warmth.ps1 (or .sh) for the SNS-engagement-on-your-posts warmth gauge.
  2. Read both people/<slug>/chat.md (private, ~70% of signal) AND people/<slug>/sns.json (public self-presentation, ~30% of signal, often flips a borderline axis)
  3. Score each of 4 axes (E/I, S/N, T/F, J/P) using signals from chat AND SNS combined (latency, voice/text ratio, abstract vs concrete language, planning style, post frequency, caption tone, audience scope, visible gaps)
  4. Walk the chat for trip-wires (reply-latency spikes, length collapse, hard redirects, meta-critique). Cross-reference with SNS for safe / unsafe topic territory and persona splits.
  5. Update people/<slug>/profile.md YAML frontmatter (and the ## 朋友圈观察 body section):
    mbti:
      type: ENFP                       # or "INFP/ENFP" if borderline, or "unclear"
      confidence: medium               # low / medium / high
      basis: |
        <2-4 sentences of specific signals>
    trip-wires:
      - topic: <short label>
        pattern: <what you did → how they reacted>
        repair: <how to avoid or recover>
    comms:
      frequency: <e.g. "1 ping / 2 days when ball is mine">
      style: <e.g. "short, no metaphors, voice notes OK">
      do: [<topic-1>, <topic-2>]
      avoid: [<thing-1>, <thing-2>]

Workflow 2 — right-now subtext read

Methodology: docs/subtext-reading.md

For every active conversation, the agent picks ONE of 6 states: 🔥 hot · 🟢 warm · 🟡 mild cool · 🟠 cooling · 🔴 disengaging · ⚫ gone

Output is a 5-line block per person, not prose:

[name]
state: 🟡 mild cool
ball: them
last move: <date> · <what happened>
key signal: <which signals matched, e.g. length collapse + soft non-answer>
do now: <one concrete action, e.g. "hold. no ping. wait 3+ days.">

Workflow 3 — task extraction from a work chat

Methodology: docs/task-extract.md

  1. First-pass filter: .\tools\task-extract.ps1 -Person <task-giver-slug> -Project <slug>
  2. Agent reads the candidate file + classifies into 6 buckets (Direct work / Decision needed / Data needed / Already-promised / Implicit / Closed)
  3. Score each task on 3 axes (stakeholder × urgency × reversibility) for P0/P1/P2
  4. Update projects/<slug>/notes.md — never wholesale rewrite, always diff
  5. Hard rule: quote, don't paraphrase. Paraphrasing introduces drift.

Workflow 4 — self-mirror

Tool: tools/self-mirror.ps1 → outputs SELF-MIRROR.md at repo root (gitignored).

Scans people/*/chat.md, auto-detects user's own sender label, produces a 7-section report: echo-reply offenses · opening-line top 10 · length distribution · question vs statement ratio · time-of-day pattern · connector/softener word frequency · emoji/sticker count. Per-recipient breakdowns throughout.

Single most actionable target: echo-reply detection ("X 看起来挺 X 啊" pattern). The 5/2 "我是在跟豆包聊天吗" call-out is the canonical failure mode this catches.

Repository structure

.
├── README.md                Human-facing intro
├── AGENTS.md                Codex / Cursor / Aider contract
├── CLAUDE.md                Claude Code workflow guide
├── SKILL.md                 ← this file (Vercel Skills entry)
├── LICENSE                  Apache-2.0
├── .gitignore               Blocks user data
├── .github/workflows/
│   └── no-data-leaked.yml   CI: blocks committing real data
├── docs/
│   ├── mbti-analysis.md
│   ├── subtext-reading.md
│   ├── task-extract.md
│   └── voice-transcription.md
├── tools/
│   ├── extract-pdf.js           PDF text extraction (Node + pdf-parse)
│   ├── contacts.ps1 / .sh       fuzzy contact lookup (verify exact name before refresh)
│   ├── refresh.ps1 / .sh        pull chat + SNS + stats for one contact (incremental)
│   ├── refresh-group.ps1 / .sh  pull a group chat into topics/<slug>/
│   ├── attachments.ps1 / .sh    list / extract chat attachments (PDFs, images, files)
│   ├── voice-transcribe.ps1 / .sh  silk → wav → whisper pipeline for voice (local-first)
│   ├── status.ps1 / .sh         one-line status per contact
│   ├── digest.ps1 / .sh         unread incremental snapshot since last session
│   ├── warmth.ps1 / .sh         SNS engagement on your posts (warmth gauge)
│   ├── self-mirror.ps1          quantify own chat habits
│   └── task-extract.ps1         first-pass TODO candidate extractor
├── people/_template/        profile.md scaffold (with full YAML schema)
├── projects/_template/      notes.md scaffold
└── topics/                  cross-chat keyword search dumps (gitignored)

Hard rules the agent must follow

  1. Never commit user data. people/<name>/, projects/<name>/, topics/<name>/ are all gitignored. CI (.github/workflows/no-data-leaked.yml) enforces.
  2. Never echo chat.md content into any context outside the local session — other chats, web searches, issue trackers, PR descriptions. Treat as password-grade.
  3. Use placeholders in any docs, commits, issues, or PRs: 张三 / 李四 / Alice / Bob / Acme Corp.
  4. Don't reimplement wx-cli. Always shell out to wx ....
  5. Match the user's communication style if stated. Sensible defaults: direct, action-oriented, no hedging, avoid echo-reply patterns when responding to the user themselves.
  6. For Codex tasks specifically: use the codex-ready label on issues you scope for Codex; the PR review checklist lives in AGENTS.md and .github/pull_request_template.md.

Dependencies

DependencyWhyInstall
@jackwener/wx-cliDecryption + querying (does all the heavy lifting)npm install -g @jackwener/wx-cli
pdf-parse (optional)For tools/extract-pdf.jsnpm install -g pdf-parse
gh (optional)For PR / issue automationhttps://cli.github.com
Node 18+, PowerShell 5.1+ or BashScriptsper-platform

Privacy & legal

Decrypt only your own WeChat data. Same rule as wx-cli. This repository is intentionally structured to make data leakage difficult: gitignore + CI + agent rules + repeated docs. If you accidentally commit data, follow GitHub's removing sensitive data guide and rotate any tokens.

Acknowledgments

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.