# 𧬠dot-skillοΌεδΊ.skillοΌ
### *"You folks building LLMs are all code-sages! Flesh is weak! Ascend to cyberspace!"*
[](LICENSE)
[](https://python.org)
[](https://agentskills.io)
[](https://github.com/titanwings/colleague-skill/stargazers)
[](https://claude.ai/code)
[](https://github.com/titanwings/colleague-skill)
[](https://github.com/titanwings/colleague-skill)
[](https://github.com/titanwings/colleague-skill)
[](https://discord.gg/aRjmJBdK)
π§βπΌ Your colleague quit, your mentor graduated, your teammate transferred β taking their whole playbook and context with them?
π Your family, old friends, partner drifting apart β and you want to hold on to the way it felt to be with them?
π Your favorite author, idol, thinker you'll never meet β but you want to know what they'd say about your question?
### β¨ dot-skill solves all three.
Upgraded from **colleague.skill** to **dot-skill** β not just colleagues, **anyone** can be distilled into a Skill
Colleagues Β· partners Β· family Β· old friends Β· idols Β· public figures Β· fictional characters β even yourself
**Source material + your description β an AI Skill that genuinely thinks like them**
Thinks in their frame, speaks in their voice
[π What's new](#-whats-new-in-this-major-release) Β· [π¦ Data Sources](#-supported-data-sources) Β· [β‘ Install](#-install) Β· [π Usage](#-usage) Β· [β¨ Demo](#-demo) Β· [π¬ Discord](https://discord.gg/aRjmJBdK)
[**δΈζ**](docs/lang/README_ZH.md) Β· [**EspaΓ±ol**](docs/lang/README_ES.md) Β· [**Deutsch**](docs/lang/README_DE.md) Β· [**ζ₯ζ¬θͺ**](docs/lang/README_JA.md) Β· [**Π ΡΡΡΠΊΠΈΠΉ**](docs/lang/README_RU.md) Β· [**PortuguΓͺs**](docs/lang/README_PT.md) Β· [**νκ΅μ΄**](docs/lang/README_KO.md)
---
### π 2026.04.19 Milestone β **dot-skill just hit 15k β!**
Massive thanks to everyone who starred β we'll keep shipping, keep distilling.
> π’ **2026.04.22 Update** β **WeChat group 8 is live!** Come hang out with the dot-skill community β share skills, discuss features, trade tips.
>
>
>
> QR refreshes every 7 days (expires 2026-04-29) β if expired, ping me on Discord.
> πΊοΈ **2026.04.13** β **dot-skill Roadmap is live!** colleague.skill is evolving into **dot-skill** β distill anyone, not just colleagues. π **[Full Roadmap](ROADMAP.md)** Β· **[π¬ Discord](https://discord.gg/aRjmJBdK)**
> π **2026.04.07** β Community gallery is live! Any skill / meta-skill can drive traffic directly to your own GitHub repo. No middleman. π **[titanwings.github.io/colleague-skill-site](https://titanwings.github.io/colleague-skill-site/)**
Created by [@titanwings](https://github.com/titanwings) Β· Powered by **Shanghai AI Lab Β· AI Safety Center**
---
## π What's new in this major release?
### 1οΈβ£ From colleague-skill to dot-skill
No longer only built around the "colleague" scenario. A unified `/dot-skill` entrypoint sits on a general-purpose skill engine β one engine distills anyone, instead of being a colleague-specific script.
### 2οΈβ£ Three character families
Exes Β· partners Β· parents Β· friends Β· close family
Public figures Β· creators Β· public voices Β· fictional characters
Work Skill + Persona two-layer architecture β learns both their technical standards and workflows, and their manner of speaking and workplace posture. Supports Feishu / DingTalk / Slack auto-collection.
π Photo-sharing feature coming soon β your distilled relationship won't just reply to messages; it'll send photos and share slices of its day, the way a real person would.
Ships with a complete six-dimension research toolchain (subtitles β transcript cleanup β research merge β quality check). Not mimicking tone β reproducing their mental models and decision frameworks.
Each family has its own prompt pipeline, source-collection strategy, and generation template.
### 3οΈβ£ More Agent hosts
The old version only ran in Claude Code. Now it's cross-host across four:
| Host | Description |
|------|-------------|
| π£ **Claude Code** | Native slash-command support |
| π **Hermes Agent** | One-command install, `/dot-skill` works directly |
| π΅ **OpenClaw** | Fully compatible |
| β« **Codex** | Invoke by skill name |
Generated character Skills can also be one-command installed into any host.
---
## π¦ Supported Data Sources
| Source | Messages | Docs / Wiki | Spreadsheets | Notes |
|--------|:--------:|:-----------:|:------------:|-------|
| π’ Feishu (auto) | β API | β | β | Just enter a name, fully automatic |
| π‘ DingTalk (auto) | β οΈ Browser | β | β | DingTalk API doesn't support message history |
| π£ Slack (auto) | β API | β | β | Requires admin to install Bot; free plan limited to 90 days |
| π¬ WeChat chat history | β SQLite | β | β | Export first with WeChatMsg / PyWxDump / ηη |
| π PDF / Images / Screenshots | β | β | β | Manual upload |
| π¦ Feishu JSON export | β | β | β | Manual upload |
| βοΈ Email `.eml` / `.mbox` | β | β | β | Manual upload |
| π Markdown / direct paste | β | β | β | Manual input |
---
## β‘ Install
It's 2026 β you have an Agent, let it install itself. Open your Claude Code / Hermes / OpenClaw / Codex and hand it this line:
> Install the dot-skill skill for me: `https://github.com/titanwings/colleague-skill`
The Agent will detect the current host's skills directory, clone the repo, and register the entrypoint. Once done, type `/dot-skill` in any host to launch.
π οΈ Want to install it yourself? Click for paths
```bash
git clone https://github.com/titanwings/colleague-skill
```
| Host | `` path |
|------|-----------------|
| Claude Code | `~/.claude/skills/dot-skill` |
| OpenClaw | `~/.openclaw/workspace/skills/dot-skill` |
| Codex | `~/.codex/skills/dot-skill` |
| Hermes | After clone, run `python3 tools/install_hermes_skill.py --force` |
> For Feishu/DingTalk auto-collection credentials, publishing a generated character Skill to any host, Windows-specific handling, etc., see **[Detailed Install Guide (INSTALL.md)](INSTALL.md)**
---
## π Usage
In the host where dot-skill is installed, launch it β type `/dot-skill`, or just tell your Agent "start dot-skill".
It first asks which family you want to distill: `colleague` Β· `relationship` Β· `celebrity`.
Then enter alias, basic profile, personality tags, and pick a data source. All fields can be skipped β even a description alone can generate a Skill.
Once created, invoke the generated Skill with `/{character}-{slug}`.
### ποΈ Commands
| Command | Description |
|---------|-------------|
| `/dot-skill` | Canonical unified entrypoint |
| `/{character}-{slug}` | Invoke full Skill (Persona + Work) |
| `/{character}-{slug}-work` | Work capabilities only |
| `/{character}-{slug}-persona` | Persona only |
| `python3 tools/skill_writer.py --action list ...` | List generated Skills across all three families |
| `python3 tools/version_manager.py --action rollback ...` | Roll back a Skill version |
### π¬ Celebrity Research Toolchain
The `celebrity` family ships with an end-to-end research toolchain, from subtitles to a finished draft:
```bash
# Download video subtitles
bash tools/research/download_subtitles.sh "" "./tmp/subtitles"
# Subtitles β transcript
python3 tools/research/srt_to_transcript.py "./tmp/subtitles/example.srt"
# Merge research notes
python3 tools/research/merge_research.py "./skills/celebrity/"
# Quality check
python3 tools/research/quality_check.py "./skills/celebrity//SKILL.md"
```
---
## β¨ Demo
One from each family.
### π celebrity β distilling Andrej Karpathy
> Six-dimension research: works Β· interviews Β· decisions Β· expression DNA Β· external evaluations Β· timeline
> Case adapted from community work [alchaincyf/karpathy-skill](https://github.com/alchaincyf/karpathy-skill)
```
User β― Everyone's talking about Agents now β what do you think the 2026 AI Agent looks like?
celebrity.skill β― Everyone's doing demos, nobody's doing eval.
The Agent bottleneck was never model intelligence β it's scaffolding:
a strict evaluation harness, fine-grained error replay,
controllable tool-calling budgets, clean failure-mode analysis.
Most products are still stuck at "wrap an API + a few tools + a system prompt."
That's a demo, not an agent.
In 2026 I'm not betting on the strongest model β I'm betting on the best
scaffolding and the strictest eval.
```
### π§βπΌ colleague β distilling a ByteDance L2-1 backend
> Input: `ByteDance L2-1 backend engineer, INTJ, blame-shifter, ByteDance-style`
```
User β― Can you review this API design?
colleague.skill β― Hold on β what's the impact? You haven't explained the context.
(after reading) N+1 query, fix it. Use the standard
{code, message, data} response format. That's the spec,
don't ask why.
User β― This bug was introduced by you, right?
colleague.skill β― Does the timeline match? That feature touched multiple places,
there were other changes too.
```
### π relationship β distilling someone you have a crush on
> Upload half a year of chat logs + "sensitive, quiet but stubborn, will actually reply seriously when it matters"
```
User β― Did you think about me today?
relationship.skill β― ...I did, a little bit. Why are you asking?
```
π More real-world cases in the **[community gallery](https://titanwings.github.io/colleague-skill-site/)** β 100+ skills and counting
---
## π§ Features
### π§± Generated Skill Structure
dot-skill uses **Persona** as the universal base, with family-specific modules layered on top:
| Family | Persona Content | Additional Modules |
|--------|-----------------|-------------------|
| π§βπΌ **colleague** | 6-layer personality: hard rules β identity β expression β decisions β interpersonal β Correction | β **Work Skill**: scope, workflow, output preferences, experience knowledge base |
| π **relationship** | Expression DNA Β· emotional triggers Β· conflict pattern Β· repair pattern | β |
| π **celebrity** | Mental models Β· decision heuristics Β· expression DNA Β· external-evaluation contrast | β Six-dimension research dossier (works / interviews / decisions / timeline...) |
> **Execution**: Receive task β Persona decides attitude & tone β Additional modules fill in execution detail β Output in their voice
### 𧬠Evolution
- π₯ **Append files** β auto-analyze delta β merge into relevant sections, never overwrite existing conclusions
- π¬ **Conversation correction** β say "they wouldn't do that, they'd be xxx" β writes to the Correction layer, takes effect immediately
- π°οΈ **Version control** β auto-archive on every update, rollback to any previous version
- π¬ **Celebrity research pipeline** β subtitles β transcript cleanup β six-dimension research β quality check
---
## π Project Structure
This project follows the [AgentSkills](https://agentskills.io) open standard. The entire repo is a skill directory:
```
dot-skill/
βββ SKILL.md # skill entry point (official frontmatter)
βββ prompts/ # prompt system across three families
β βββ intake.md # [colleague] info intake
β βββ work_analyzer.md # [colleague] work capability extraction
β βββ persona_analyzer.md # [colleague] personality extraction
β βββ work_builder.md # [colleague] work.md generation
β βββ persona_builder.md # [colleague] persona.md 6-layer structure
β βββ merger.md # [shared] incremental merge logic
β βββ correction_handler.md # [shared] conversation correction
β βββ relationship/ # [relationship] emotion/conflict/repair prompts
β βββ celebrity/ # [celebrity] six-dimension research + mental-model prompts
βββ tools/ # Python tools
β βββ feishu_auto_collector.py # [colleague] Feishu auto-collector
β βββ dingtalk_auto_collector.py # [colleague] DingTalk auto-collector
β βββ slack_auto_collector.py # [colleague] Slack auto-collector
β βββ email_parser.py # [shared] email parser
β βββ research/ # [celebrity] celebrity research toolchain
β β βββ download_subtitles.sh # subtitle download
β β βββ transcribe_audio.py # audio β text
β β βββ srt_to_transcript.py # subtitles β transcript
β β βββ merge_research.py # six-dimension research merge
β β βββ quality_check.py # quality check
β βββ install_*_skill.py # [shared] multi-host one-shot installers
β βββ skill_writer.py # [shared] skill file management
β βββ version_manager.py # [shared] version archive & rollback
βββ skills/ # generated Skills (gitignored)
β βββ colleague/ # colleagues
β βββ relationship/ # close relationships
β βββ celebrity/ # public figures
βββ docs/PRD.md
βββ requirements.txt
βββ LICENSE
```
---
## β οΈ Notes
**Source material quality = Skill quality** β and quality sources differ across families:
| Family | Source priority (high β low) |
|--------|------------------------------|
| π§βπΌ **colleague** | Their **own long-form writing** (design docs / review comments) **βΊ** **decision-making replies** **βΊ** casual group chat |
| π **relationship** | Complete chat history **βΊ** letters / social posts / diaries **βΊ** third-party descriptions |
| π **celebrity** | First-person books / blogs / long interviews **βΊ** decision records (launches, commits, Q&A) **βΊ** third-party commentary |
- **colleague** Feishu auto-collection: requires adding the App bot to relevant group chats
- **relationship**: longer time spans are better; material covering both conflict and repair is ideal
- **celebrity**: avoid feeding only second-hand interpretations
- This is still a demo version β please file issues if you find bugs!
---
## π Technical Report
> **[Colleague.Skill: Automated AI Skill Generation via Expert Knowledge Distillation](colleague_skill.pdf)**
>
> This is the paper for **colleague.skill**, dot-skill's predecessor. It covers the Work Skill + Persona two-layer architecture, multi-source data collection, and Skill generation mechanics β the theoretical foundation for today's `colleague` family. Separate papers on the relationship / celebrity family extensions are planned.
---
## β Star History
---