Skills
TL;DR. Skills are narrow, one-task playbooks for AI agents. Each is a Markdown file under
skills/in the repo. Agents using the MCP server can also pull them aswun://skills/<name>resources on demand.
The skills/ directory holds narrow, single-task how-to playbooks
agents can ingest and follow mechanically. Each skill has the same
shape:
# Task name
## When to use this – concrete signals this is the right skill
## Inputs you need – what the agent should ask for
## Steps – a numbered list, ~5–10 items max
## Verify – commands or behaviour confirming successIf a task takes more than ~10 steps, it should be split.
Available skills
| skill | when to use |
|---|---|
| add-screen-with-form | adding a screen that takes user input |
| wire-an-intent | shipping a new intent end-to-end |
| add-component-pack | building a reusable :myapp/* library |
| ship-a-breaking-change | landing a change that needs a migration script |
Each skill is also a MCP resource under
wun://skills/<name>, so agents using the wun-mcp server can pull
them into context on demand.
Adding a skill
Drop a new .md file under skills/ matching the shape above.
Keep them narrow — one task per file. The
skills README
in the repo has the full rubric.