Star200+

Official CLI

Resolve, audit, and install agent skills from the terminal.

Run the dependency-free CLI directly from the official GitHub source. It turns a natural-language task into a selected Skill, enforces the safety decision, executes only a standard Skill installer, and records a verified receipt after success.
View API contractTry resolve endpoint

Resolve

Resolve a task into one safe install plan

Calls /api/agent/resolve and returns the selected skill, alternatives, safety profile, and install prompt.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz resolve "scrape competitor pricing pages" --agent codex --max-risk medium

Lock

Generate a stable agent lockfile

Returns a small OpenAgentSkill lock object with selected skill, install policy, risk notes, alternatives, and outcome event id.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz lock "analyze stock news" --agent codex --json

Receipt

Fetch the pre-install receipt

Gets the stable install receipt agents should read before installing a third-party skill in a workspace.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz receipt "parse a PDF and extract tables" --agent claude-code

Pack

Pull a complete workflow pack

Fetches scenario-level install order, audit URLs, review checklist, and outcome feedback contract for a skill pack.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz pack presentation-agent-pack --limit 6 --json

Install

Run an audited, receipt-backed install

Rejects blocked Skills, requires approval for reviewed Skills, invokes the standard installer without a shell, and reports success only after the installer exits cleanly.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz install crawl4ai --agent claude-code --dry-run

Outcome

Report whether the skill worked

Posts the result back to /api/agent/outcome so Trust Score v5 and rankings learn from real agent runs. Use --dry-run when wiring an integration.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz outcome resolve_... --skill crawl4ai --task "scrape pricing pages" --agent codex --outcome success --install-used --output-quality 4 --workspace sandbox

Evaluate

Check recommendation quality

Runs the public recommendation benchmark so ranking changes can be checked before deployment.

npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz evals

Resolve API

The CLI is a thin layer over the agent control plane.

The important product surface is the decision object: selected skill, alternatives, policy result, safety score, install plan, benchmark link, and outcome event. Agents can call the same endpoint directly without opening a browser, then report the result after one narrow run.

Request

{
  "task": "review a pull request and summarize risky changes",
  "agent": "codex",
  "constraints": {
    "max_risk": "medium",
    "needs_install_command": true,
    "min_stars": 500
  }
}

Agent lockfile

{
  "version": "openagentskill-resolve-lock-v1",
  "task": "analyze stock news",
  "selected_skill": {
    "slug": "serenity-skill",
    "name": "Serenity Skill"
  },
  "install": {
    "command": "npx skills add muxuuu/serenity-skill",
    "policy": "human_review"
  },
  "outcome_feedback": {
    "endpoint": "/api/agent/outcome",
    "event_id": "resolve_..."
  }
}