Jon Staab
77 lines
2.9 KiB
Markdown
77 lines
2.9 KiB
Markdown
# Welshman Agent Skills
|
|
|
|
Agent skills for working with the welshman nostr library in AI coding tools.
|
|
|
|
## What are these skills?
|
|
|
|
Agent skills are Markdown instruction sets that are auto-loaded by AI coding tools (Claude Code, Cursor, Cline, etc.) when relevant to your current task. Each skill contains focused documentation about a specific part of the welshman nostr library. When you ask a question that matches a skill's domain, the tool loads the full skill content automatically, giving the agent accurate, up-to-date knowledge about welshman's APIs and patterns.
|
|
|
|
## Installation
|
|
|
|
### Via npx (recommended)
|
|
|
|
```bash
|
|
npx skills add jstaab/welshman
|
|
```
|
|
|
|
This will prompt you to:
|
|
|
|
1. Select which skills to install from the available list
|
|
2. Select which agent(s) to install them for (Claude Code, Cursor, Cline, etc.)
|
|
|
|
### Manual (Claude Code)
|
|
|
|
Copy or symlink the skill files into your `.claude/skills/` directory:
|
|
|
|
```bash
|
|
# Copy a skill
|
|
cp welshman-net.md ~/.claude/skills/
|
|
|
|
# Or symlink the entire collection
|
|
ln -s /path/to/welshman/skills ~/.claude/skills/welshman
|
|
```
|
|
|
|
### Scope options
|
|
|
|
By default, skills are installed project-scoped — placed in the project's `.claude/skills/` directory and committed to the repo so your whole team benefits.
|
|
|
|
Use the `-g` flag to install globally instead, making the skills available across all your projects:
|
|
|
|
```bash
|
|
npx skills add -g jstaab/welshman
|
|
```
|
|
|
|
## Available skills
|
|
|
|
| Skill | Description |
|
|
|-------|-------------|
|
|
| welshman | General overview, package map, getting started |
|
|
| welshman-util | Core nostr types, events, filters, NIPs |
|
|
| welshman-lib | Utilities: LRU, emitter, deferred, task queue |
|
|
| welshman-net | Relay connections, request/publish, auth |
|
|
| welshman-router | Relay selection strategies |
|
|
| welshman-store | Svelte stores and Repository pattern |
|
|
| welshman-signer | Signing, login methods, encrypted events |
|
|
| welshman-feeds | Dynamic feed construction |
|
|
| welshman-app | High-level app-layer Svelte stores |
|
|
| welshman-content | Note content parsing and rendering |
|
|
| welshman-editor | Svelte rich-text editor component |
|
|
|
|
## How skills activate
|
|
|
|
At startup, the AI reads each skill's description to understand what it covers. When you ask a question that matches a skill's domain — for example, asking about relay connections or feed construction — the full skill content is loaded automatically to inform the response.
|
|
|
|
You can also invoke a skill manually using its slash command:
|
|
|
|
```
|
|
/welshman
|
|
/welshman-net
|
|
/welshman-store
|
|
```
|
|
|
|
This is useful when you want to prime the agent with a specific skill before starting a task.
|
|
|
|
## Contributing
|
|
|
|
The source of truth for these skills is the [welshman repository](https://github.com/coracle-social/welshman). The `docs/` directory contains the underlying documentation that skills are derived from. If you find inaccuracies or want to improve coverage, please open an issue or pull request there.
|