How do I install the Hart Skill Authoring skill?

Run npx skillmds@latest add javimosch/hart-skill-authoring in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Hart Skill Authoring safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Hart Skill Authoring?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Hart Skill Authoring free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Hart Skill Authoring?

javimosch (@javimosch) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Hart Skill Authoring by javimosch — Agent Skill

Name: Hart Skill Authoring
Author: javimosch

File contents

hart Skill Authoring

Publish agent skills as hart artifacts so they are versioned, gated, and discoverable.

When to use

The skill is a runbook, recipe, or one-shot fix.
You want a stable URL, versioning, or password gating.
The skill does not need a fancy report/dashboard layout.

Artifact format

A hart skill is a plain HTML file with no CSS, no external resources, and a semantic body.

Required <head>:

<meta charset="utf-8">
<meta name="title" content="Skill title">
<meta name="description" content="One-line description">
<meta name="keywords" content="foo, bar, baz">
<title>Skill title</title>

Required structure:

<!doctype html>, <html>, <head>, <body>.
<h1> for the title.
<h2> for sections.
<ul> / <ol> for lists.
<pre><code> for commands.
No <style>, no <link>, no <script>, no images, no external URLs.

Keep the whole file under 200 lines.

Minimal template

<!doctype html>
<html lang="en">
<head>
  <meta charset="utf-8">
  <meta name="title" content="Fix X">
  <meta name="description" content="Recover from X by doing Y">
  <meta name="keywords" content="x, y, z">
  <title>Fix X</title>
</head>
<body>
  <h1>Fix X</h1>
  <p>One-line trigger.</p>

  <h2>Prerequisites</h2>
  <ul>
    <li>Thing 1</li>
  </ul>

  <h2>Steps</h2>
  <pre><code>command one
command two</code></pre>

  <h2>Verify</h2>
  <pre><code>verify command</code></pre>
</body>
</html>

Publish

hart publish fix-x.html \
  --owner jar-skills \
  --artifact fix-x \
  --title "Fix X" \
  --visibility private \
  --read-key gtf \
  --tags x,y,z \
  --meta '{"trigger":"claude failed with X","verified":true}'

Use --visibility unlisted if no password is needed.

Verify

wc -l fix-x.html          # should be < 200
hart publish fix-x.html --dry-run
hart list --owner jar-skills --format skills  # confirm the skill appears in the catalog
claude -p "read https://hart.intrane.fr/a/jar-skills/fix-x and follow it"

Notes

Description/keywords are read from <meta> tags and exposed automatically in hart list --owner <who> --format skills and GET /v1/skills/<owner>.
For password-gated skills, agents fetch with X-Hart-Read-Key: <password> or HART_READ_KEY.
Re-publish the same --owner/--artifact to version the skill.

---name: hart-skill-authoringdescription: Publish agent skills as minimal, discoverable hart HTML artifacts using meta title/description/keywords.---# hart Skill AuthoringPublish agent skills as hart artifacts so they are versioned, gated, and discoverable.## When to use- The skill is a runbook, recipe, or one-shot fix.- You want a stable URL, versioning, or password gating.- The skill does not need a fancy report/dashboard layout.## Artifact formatA hart skill is a plain HTML file with no CSS, no external resources, and a semantic body.Required `<head>`:```html<meta charset="utf-8"><meta name="title" content="Skill title"><meta name="description" content="One-line description"><meta name="keywords" content="foo, bar, baz"><title>Skill title</title>```Required structure:- `<!doctype html>`, `<html>`, `<head>`, `<body>`.- `<h1>` for the title.- `<h2>` for sections.- `<ul>` / `<ol>` for lists.- `<pre><code>` for commands.- No `<style>`, no `<link>`, no `<script>`, no images, no external URLs.Keep the whole file under 200 lines.## Minimal template```html<!doctype html><html lang="en"><head> <meta charset="utf-8"> <meta name="title" content="Fix X"> <meta name="description" content="Recover from X by doing Y"> <meta name="keywords" content="x, y, z"> <title>Fix X</title></head><body> <h1>Fix X</h1> <p>One-line trigger.</p> <h2>Prerequisites</h2> <ul> <li>Thing 1</li> </ul> <h2>Steps</h2> <pre><code>command onecommand two</code></pre> <h2>Verify</h2> <pre><code>verify command</code></pre></body></html>```## Publish```bashhart publish fix-x.html \ --owner jar-skills \ --artifact fix-x \ --title "Fix X" \ --visibility private \ --read-key gtf \ --tags x,y,z \ --meta '{"trigger":"claude failed with X","verified":true}'```Use `--visibility unlisted` if no password is needed.## Verify```bashwc -l fix-x.html # should be < 200hart publish fix-x.html --dry-runhart list --owner jar-skills --format skills # confirm the skill appears in the catalogclaude -p "read https://hart.intrane.fr/a/jar-skills/fix-x and follow it"```## Notes- Description/keywords are read from `<meta>` tags and exposed automatically in `hart list --owner <who> --format skills` and `GET /v1/skills/<owner>`.- For password-gated skills, agents fetch with `X-Hart-Read-Key: <password>` or `HART_READ_KEY`.- Re-publish the same `--owner/--artifact` to version the skill.

Frequently asked questions

Hart Skill Authoring