Skill Publisher

Packages a local skill directory into a Shipables-ready archive.

Workflow

Step 1 — Locate the skill

Ask the user for the path to their skill directory (e.g. ~/.claude/skills/my-skill). If they've already provided it, skip asking.

Verify the directory exists and contains a SKILL.md. If not, tell the user and stop.

Step 2 — Parse SKILL.md frontmatter

Read the SKILL.md file and extract the YAML frontmatter fields:

• name — expected format is Author/SkillName (e.g. UJameel/Sift)
• description — the one-line description string

Split name on / to get:

• author = the part before / (e.g. UJameel)
• skill_slug = the part after /, lowercased (e.g. sift)

If name has no /, use the whole value as skill_slug and prompt for the author separately.

Step 3 — Collect missing metadata interactively

Prompt the user for the following, one at a time or as a grouped list. Use sensible defaults shown in brackets:

1. Version [1.0.0] — what version is this release?
2. License [MIT] — what license applies?
3. GitHub repo URL — used for both homepage and repository. Format: https://github.com/Author/SkillName. Offer to auto-construct it from the parsed author and skill_slug (e.g. https://github.com/UJameel/Sift) — ask user to confirm or correct.
4. Tags — suggest 4–6 tags auto-generated from the skill's description and body content (lowercase, hyphenated slugs). Ask user to confirm, add, or remove.

Step 4 — Generate shipables.json

Create shipables.json with this exact structure:

{
  "name": "<skill_slug>",
  "version": "<version>",
  "description": "<description from frontmatter>",
  "author": "<author>",
  "license": "<license>",
  "tags": ["<tag1>", "<tag2>", "..."],
  "homepage": "<github url>",
  "repository": "<github url>"
}

Write this file to a temp working directory (e.g. /tmp/skill-publisher-<skill_slug>/).

Copy the skill directory contents there too:

cp -r <skill_dir>/. /tmp/skill-publisher-<skill_slug>/

Then write shipables.json into that temp directory.

Step 5 — Create the archive

Run the following from inside the temp directory, including only files/folders that actually exist:

cd /tmp/skill-publisher-<skill_slug>

# Build the include list dynamically
INCLUDE="SKILL.md shipables.json"
[ -d scripts ]    && INCLUDE="$INCLUDE scripts/"
[ -d references ] && INCLUDE="$INCLUDE references/"
[ -d assets ]     && INCLUDE="$INCLUDE assets/"
[ -d agents ]     && INCLUDE="$INCLUDE agents/"

tar -czf /tmp/<skill_slug>.tar.gz $INCLUDE

Step 6 — Report to user

Tell the user:

Your skill archive is ready:

  /tmp/<skill_slug>.tar.gz

Go to Shipables → Publish a Skill, enter your skill name as:

  <author> / <skill_slug>

Then drag and drop the archive file into the upload area and click Publish Skill.

Also show a summary of what was included in the archive (list the files/folders bundled).

Notes

• Never overwrite an existing published version without confirming with the user.
• If shipables.json already exists in the skill directory, ask the user if they want to regenerate it or use the existing one.
• Tags should be meaningful and discoverable — draw them from key tools, domains, or use cases mentioned in the skill body. Avoid generic tags like "skill" or "claude".