source.mda

name	mda-compat-intro
description	MDA compatibility demonstration skill for Claude Code. Compiled from compat/claude-code/source.mda. Loading this skill confirms that an MDA-emitted SKILL.md loads in Claude Code with the envelope subset of agentskills.io v1 fields honored.
allowed-tools	Read
metadata	mda doc-id title version tags 51ba1f80-1e9a-4f3c-8b6e-8a4c1b2d3e4f MDA Compat Demo (Claude Code) 1.0.0 compat demo claude-code

# MDA Compat Demo (Claude Code) This skill exists to verify Claude Code SKILL.md compatibility from an MDA-compiled output. Three things are true if you can see this skill listed in your Claude Code session: 1. The compiled `SKILL.md` parses under Claude Code's loader. 2. The frontmatter `name` matches the package directory name (spec §06-2.1). 3. The top-level `allowed-tools` field reaches Claude Code where it can be honored. Source: `compat/claude-code/source.mda`. Compile: `mda compile compat/claude-code/source.mda --target SKILL.md --out-dir compat/claude-code/build --integrity`. What this does NOT prove: - That MDA today can author the runtime-specific Claude Code fields outside the agentskills.io v1 envelope (`when_to_use`, `model`, `effort`, `context`, `hooks`, `paths`, `shell`, `agent`, `argument-hint`, `arguments`, `disable-model-invocation`, `user-invocable`). Those are out of v1.0 scope and earmarked for v1.1. - That Claude Code parses MDA-extended frontmatter under `metadata.mda.*` (it does not, and is not required to). - That signatures or integrity digests are checked at load (no current runtime does that).