The simplest useful Claude-compatible skill is a folder with a SKILL.md file. The portable Agent Skills format makes only two frontmatter fields mandatory: name and description. The value comes from what you put below the frontmatter: the workflow, examples, validation, and pointers to supporting files.

Copyable SKILL.md template

---
name: release-readiness
description: Reviews release readiness for static websites, including build output, sitemap, robots.txt, source citations, deployment verification, and final report requirements. Use before publishing or auditing a static site.
license: MIT
compatibility: Designed for coding agents with shell, file, and browser access.
---

# Release Readiness

## When to Use
Use this skill when the user asks to:
- Prepare a static site for launch.
- Audit generated SEO, GEO, and deployment artifacts.
- Build, deploy, and verify a website release.

Not for general design feedback unless release readiness is in scope.

## Workflow
1. Read the brief and source-of-truth docs before editing.
2. Inspect the current site structure, build scripts, and generated artifacts.
3. Make the smallest complete changes needed for the release gate.
4. Run the project build and inspect the output.
5. Deploy only after build success, then verify live URLs.
6. Record checks, skips, and deployment evidence in the final report.

## References
- references/release-checklist.md - detailed checklist.
- assets/report-template.md - final report outline.

## Validation
- Run npm run build.
- Confirm sitemap, robots.txt, canonical tags, OG image, and JSON-LD.
- Verify the deployed homepage and one deep page return HTTP 200.
- Report any skipped check with the reason.

Why each section exists

The frontmatter is the discovery surface. The name must be a stable lowercase identifier, and the description should name both the capability and trigger conditions. The body is the operating procedure the agent reads after activation.

The When to Use section reduces accidental activation. The Workflow section carries the ordered steps. The References section tells the agent which supporting files exist and when they matter. The Validation section defines success so the skill does not end at "I changed something."

Common template mistakes

  • Using a broad description such as "helps with websites" instead of trigger-rich wording.
  • Letting the skill name differ from the folder name.
  • Putting long policy manuals in SKILL.md instead of referenced files.
  • Naming scripts or assets that are not bundled with the skill.
  • Skipping validation steps, which makes the workflow hard to trust.

Make one now

Use the Skill Scaffold Generator to create a first draft, then paste the description into the Description Linter. If the workflow might need live APIs, state, or auth, run the Skill vs MCP Decider before overloading the skill with integration logic.

For the larger design model, read Anatomy of a skill, Authoring reusable skills, and How portable are skills?.