# Memeroot · Skill Author

**Start here:** open `GUIDE.html` in a browser.

A working demonstration of Claude Code skills as cryptographically signed, transferable artifacts. Walk through authoring a skill as a Memeroot region, signing it as the author, and bundling it as one HTML file the recipient verifies and installs.

## What this stream demonstrates

Claude Code skills today live as folders with SKILL.md (YAML frontmatter + markdown). Sharing is via repo, copy-paste, or manual transfer; trust is implicit; modifications are invisible; lineage breaks at every copy.

This stream demonstrates a different shape: skills as Memeroot artifacts that carry their authorship cryptographically. Each skill is signed by its author; the signature is mathematically verifiable; tampering is detected; forks carry lineage; bundles transfer the whole assembly with verification intact.

The recipient opens the bundle.html in any browser, sees the skill rendered as documentation, clicks the signature to verify, copies the rendered SKILL.md block into their Claude Code skills folder. No marketplace operator; the cryptography distributes trust.

## What's in this zip

```
memeroot-skill-author/
├── GUIDE.html              ← open this first (15-step interactive guide)
├── README.md               ← you are here
├── canvas/MR-CANVAS-v0.8.html
├── features/               ← ten feature XMLs
├── identity/identity-tour-author.json
├── stylesheets/
│   ├── xsl-tone-{terse,public,annotated}.xml
│   └── xsl-skill-doc.xml      ← renders <skill> as docs + SKILL.md export
├── examples/
│   └── skill-example.xml      ← regulated-document-author skill
└── stream/
    ├── stream-skill-author.xml
    └── chunk-skill-1.xml … chunk-skill-6.xml
```

## The example skill

The stream uses a worked example: a `regulated-document-author` skill that helps Claude Code generate SOX controls, FDA records, IRB protocols, and similar compliance documentation. The choice connects to the compliance-audit-trail stream — skill provenance matters most where the skill's outputs face regulatory scrutiny. A compliance team installing this skill from an unknown source wants confidence about who wrote the instructions Claude will follow.

The example skill teaches six concrete instructions for regulated-document authoring:
1. Identify the regulatory frame (SOX, FDA Part 11, ICH GCP, HIPAA, DFARS, ...)
2. Structure as addressable regions (not monolithic documents)
3. Apply separation of duties to attestations
4. Preserve fork lineage on revisions
5. Bundle as the deliverable, not the platform record
6. Verify signatures before consuming external attestations

Customizable via text editor + re-paste. The XML shape works for any skill domain.

## The trust topology this enables

A skill ecosystem with cryptographic provenance creates a different shape of market:

- **Skill packs from domain experts** — a consultancy publishes signed bundles for verticals (financial controls, healthcare compliance, legal review). Each skill independently verifiable.
- **Regulated-skill assurance** — a regulator or compliance body maintains a list of approved skill-author identities. Skills signed by those identities are pre-approved for use in regulated contexts.
- **Forks with lineage** — a user customizes a base skill, signs the fork, distributes. The fork-source element ties their version back to the original.
- **Pay-gated distribution** — skill bundle distributable freely; SKILL.md content encrypted (vault feature); decryption requires payment-derived key. No DRM platform needed.
- **Multi-language attribution** — author writes in native language; translator counter-signs translation; consumers see both signatures.

What's structurally common to all these: cryptography distributes trust so no central operator is needed. The bundle distributes the artifact so no service is needed. Signing distributes authorship so no platform identity is needed.

## Install path for the recipient

After the recipient opens your bundle.html and verifies:
1. Copy the SKILL.md code block from the rendered skill
2. Create `~/.claude/skills/{skill-name}/` (or wherever their Claude Code setup expects skills)
3. Paste the content as `SKILL.md` in that folder
4. Restart Claude Code (or wait for next session)

Claude picks up the skill and invokes it when the task description matches the activation description in the skill's frontmatter.