LIUSIR Skills

AI agent skills for source collection, content workflows, and publishing automation.

This repository currently includes three workflow skills:

Local media and article collection workflow
WeChat Official Account article workflow
Xiaohongshu note workflow

Available Skills

Skill	Description
ls-multi-collector	Collect Douyin, YouTube, WeChat, X, and generic web content into local bundles with video download, transcript, and article-fetch workflows
ls-wechat-article	Write and publish WeChat Official Account articles end-to-end — topic intake, drafting, SEO polish, cover and inline images, theme preview, draft publish, stats backfill, and learning workflows
ls-xhs-note	Create Xiaohongshu note assets end-to-end — topic intake, source article drafting, native note writing, visual planning, and image generation

Quick Install

Install a specific skill

npx skills add FrankL1u/liusir-skills --skill ls-multi-collector
npx skills add FrankL1u/liusir-skills --skill ls-wechat-article
npx skills add FrankL1u/liusir-skills --skill ls-xhs-note

See available skills

npx skills add FrankL1u/liusir-skills --list

Install from a marketplace plugin

This repository also includes a plugin-style distribution path through .claude-plugin/.

Prerequisites

Some skills may require additional local setup.

Common prerequisites across the workflow skills:

Node.js >= 18
Python >= 3.9
ls-multi-collector also requires uv, ffmpeg, yt-dlp, defuddle, xreach, and camoufox
Optional image provider keys for AI image generation

Skill-specific notes:

ls-multi-collector: collects source material into local bundles; remote ASR and LLM are optional for transcript enhancement
ls-wechat-article: requires WeChat Official Account API credentials for publishing, and can optionally use TrendRadar MCP for topic signals
ls-xhs-note: does not publish directly; it can optionally use TrendRadar MCP for topic signals and image providers for Step 5 generation

See the setup guides in skills/ls-multi-collector/README.md, skills/ls-wechat-article/README.md, and skills/ls-xhs-note/README.md.

Works With

These skills are designed for AI agents and coding tools that support skill-style packaging:

OpenClaw
Claude Code
Cursor
Codex
Gemini CLI
Windsurf
Kilo
OpenCode
Goose
Roo
Any tool supporting npx skills add

Repository Layout

.
├── .claude-plugin/        # Plugin-style distribution metadata
├── .github/workflows/     # CI and release workflows
├── scripts/               # Repository-level maintenance scripts
├── shared/                # Shared templates and publishing docs
└── skills/                # Installable skills

Contributing

Add a new skill

Create skills/<skill-name>/
Add a SKILL.md with name, version, description, triggers, and usage rules
Add skill-specific references, scripts, and runtime files inside that skill directory
Update any plugin or marketplace metadata if needed
Open a pull request

Update an existing skill

Make your changes
Bump the version: in SKILL.md
Update skill-specific docs if behavior changed
Open a pull request

Publish

CI can publish changed skills through the repository release flow
Repository maintenance scripts live under scripts/

Notes

Each skill should keep its own agent-facing files inside its skill directory
Runtime output, caches, secrets, and virtualenvs should stay out of skills/
shared/ is repository-level guidance and publishing support, not per-skill runtime output

License

MIT