Skip to content

Skills are small agent packages published on this site. Each skill is one
directory with a SKILL.md file plus optional scripts, references, and
assets, following the open Agent Skills standard from agentskills.io.

Every skill carries a short and a long description, both from the same
specification so every skill reads the same way.

Short: the frontmatter description in SKILL.md. One to 1024 characters
stating what the skill does and when to use it, with trigger keywords.
Agents load every installed short description at startup, and the skills
list shows exactly this text per skill.

Long: the markdown body after the frontmatter. Rendered on the skill page
below the facts table. It starts with one H1 title, then the fixed
sections What it does, When to use, Instructions, Examples, References.
Keep it concise and under 500 lines; move detail into references/ files
and deterministic work into scripts/.

The name is 1 to 64 characters of lowercase letters, digits, and
hyphens, with no leading, trailing, or consecutive hyphens, and it must
match the skill directory name. Optional fields are license,
compatibility (1 to 500 characters, only when the skill needs a
specific environment), metadata (a string map, used here for author and
version), and the experimental allowed-tools.

Admins author skills at /admin/skills: New skill writes a fresh
SKILL.md, Edit rewrites an existing one. Every save validates in the
official way: strict specification checks first, then
muse skills validate on the written directory. A save lands only when
both pass; a failed create removes its directory and a failed update
restores the previous file. Draft from prompt asks the configured model
to propose both descriptions, which the admin reviews before saving.

The skill page lists the full file tree with per-file downloads and one
ZIP holding the whole package. Build artifacts and hidden files are
excluded from both. The skills index and every skill page are indexed
and listed in the sitemap.

8.1ms