Index of every page on this site: llms.txt
All skills
google avatar

/find-skills

@ead8d7f
by googlegoogle/adk-recipes11k stars
2,961
Run it yourself

Discover/install NEW agent skills from external sources (skills.sh). Use for capabilities you lack, e.g. "find a skill for X". Do NOT use to list skills you already have; see `<available_skills>`.

  • 1 file
  • 4.2 KB
  • Updated 3 months ago
  • GitHub
Use this Skill: https://skilld.dev/gh/google/adk-recipes/find-skills

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈52 tokens for metadata: the name and description. ≈1k when used: this file.

Description uses 2.6% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈210 characters, or 2.6%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Find skills

When to use this skill

  • The user asks "find a skill for X", "is there a skill that does X"
  • The user asks "how do I do X" where X is a common task with a likely existing skill
  • The user wants to extend Horizon's capabilities with a new skill

Before searching externally, check the <available_skills> block in your context — the user may already have a relevant skill installed.

Step 1 — check the leaderboard

Before running any CLI command, check skills.sh to see whether a well-known skill already covers the domain. The leaderboard ranks by total installs, surfacing the most popular and battle-tested options.

Step 2 — search for skills

If the leaderboard doesn't cover the user's need, search via the Skills CLI:

bash(command="npx --yes skills find <query>")

Examples:

User says Search
"make my React app faster" npx --yes skills find react performance
"help with PR reviews" npx --yes skills find pr review
"create a changelog" npx --yes skills find changelog

If npx is unavailable the command will fail — fall back to browsing skills.sh manually and pasting the SKILL.md content.

Step 3 — verify quality before recommending

Never recommend a skill based solely on search results.

  • Install count — prefer 1 000+ installs; treat anything under 100 with skepticism
  • Source reputation — official organisations (vercel-labs, anthropics, microsoft) are more trustworthy than unknown authors
  • GitHub stars — check the source repo; a skill from a repo with <100 stars is a yellow flag

Step 4 — present the option to the user

I found a skill that might help: "react-best-practices" from Vercel Engineering
provides React and Next.js performance guidelines (185K installs).

Shall I install it?

Only proceed to installation after the user confirms.

Step 5 — install

Option A — write SKILL.md directly (always works)

If you have the skill content (pasted by the user, retrieved from skills.sh, or constructed from the search result):

write(".agents/skills/<name>/SKILL.md", content)
load_skill(action="reload")

write creates the directory if it does not exist. load_skill(action="reload") refreshes the <available_skills> catalog (plus extensions and manifest) on the next turn.

Option B — install via npx (requires Node.js)

First check whether npx is available:

bash(command="which npx")

If it is, install the skill. Select a single skill from a multi-skill repo with the @<skill> suffix (omit it to take the whole repo):

bash(command="npx --yes skills add <owner/repo>@<skill-name> -y")

The Skills CLI stages the skill (with its references/, assets/, and scripts/ subdirs) under .agents/skills/<skill-name>/ — exactly where Horizon loads skills from — so no move is needed. Just refresh:

load_skill(action="reload")

The staged directory name MUST match the name: field in the SKILL.md frontmatter — _load_skill_from_dir validates that and skips mismatched skills.

Step 6 — confirm activation

After load_skill(action="reload") returns (look for the new skill in loaded), tell the user:

The "<name>" skill is now installed. It will appear in your
<available_skills> catalog on the next turn.

When no skills are found

  1. Acknowledge that no existing skill was found
  2. Offer to help directly using general capabilities
  3. Offer to write a custom skill for them:
bash(command="npx --yes skills init my-skill-name")

Or skip the CLI and author one manually — write a SKILL.md following the policy or workspace skills as examples.

Common skill categories

Category Search terms
Web development react, nextjs, typescript, css, tailwind
Testing testing, jest, playwright, e2e
DevOps deploy, docker, kubernetes, ci-cd
Documentation docs, readme, changelog, api-docs
Code quality review, lint, refactor, best-practices
Design ui, ux, design-system, accessibility

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at ead8d7f. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 days ago.

Activeupdated 3 months ago

README badge

README badge for google/adk-recipes/find-skills