OpenAgentSkill CLI
A skill registry with a CLI and API: the agent describes a task and gets a recommendation with a trust score and install command
- 366 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The CLI calls an external service and may send telemetry until a flag turns it off
- After confirmation the add command installs a third-party skill into the project
Install
- Works in
Manual install
npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.3.0/openagentskill-0.3.0.tgz resolve "extract tables from PDF reports" --agent codexThe package is pinned to cli-v0.3.0 and has no dependencies. Use find to search without installing and add with --dry-run for a plan.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The CLI calls an external service and may send telemetry until a flag turns it off
- After confirmation the add command installs a third-party skill into the project
This is third-party code. Review the repository files before installing.
What it does
OpenAgentSkill is a skill registry with an interface built for the agent itself. You or the agent state a task, such as extracting tables from PDF reports, name the target (Codex, Claude Code, Cursor) and get one recommended skill plus alternatives. The recommendation comes with a Trust Score v5, a link to the audit, an install command and a receipt that can be handed on. The score combines task fit with repository evidence: install readiness, maintenance, license clarity, risk signals and real agent outcomes. There are three ways in: the Resolve web page, a request to `/api/agent/resolve` and the CLI. The API answers in JSON, plain text or a compact lockfile. The CLI package has no dependencies, refuses blocked skills and requires explicit confirmation before executing. After one narrow run the agent can report the outcome to `/api/agent/outcome`, and that data feeds later scores.
Who it is for. For developers and agent builders who install third-party skills into Codex, Claude Code or Cursor and want to compare options and vet them before installing.
Good fit when
- You do not know which of dozens of skills to pick for a job like PDF work or browser automation
- Before installing someone else's skill you need to see its license, maintenance and access level
- The agent should consult the check by itself before installing: there is a ready prompt text for that
- You want to see an install plan without executing anything
- You want to feed the result of a skill run back to the registry so scores rest on real cases
Not a fit when
- You work without network access: matching goes through the openagentskill.com service
- You need a guarantee that a skill is safe: the project gives none and calls the score an aid to the decision
- You need your own private registry: the project is public and offers no private deployments
Example request
Use OpenAgentSkill to find a skill for extracting tables from PDFs for Codex and show the install plan without running itLimitations
- Matching and audits are served by the external openagentskill.com service. The CLI does not work offline.
- The score helps choose a skill but does not confirm third-party code is safe. The author advises running a skill in a sandbox first and reporting after one narrow run.
- The CLI ships as a release archive and is not published to npm. Commands are pinned to cli-v0.3.0 and run the archive by its address through `npx --yes`.
- A TypeScript SDK sits in packages/sdk and is not yet published to npm.
- Telemetry is turned off with `--no-telemetry`, and the CLI documentation lists what is sent.
How to disable. The CLI has no permanent install: it runs through `npx --yes` and registers nothing. Remove skills installed through it the usual way for your agent.
FAQ
How do I run the OpenAgentSkill CLI?
Run `npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.3.0/openagentskill-0.3.0.tgz resolve "extract tables from PDF reports" --agent codex`. The `find` command searches without installing, and `outcome-contract` prints the outcome reporting rules.
What does a resolve request return?
The best skill, ranked alternatives, a Trust Score v5 with a decision, an audit link, the risk level, an install command and a receipt. The same request is available as JSON, plain text and a compact lockfile.
How do I give an agent the API request?
Send a GET to `/api/agent/resolve` with task, agent, max_risk and format parameters. Ready templates for Codex, Claude Code and Cursor are in the Agent Integration Kit on the site.
What does the CLI do with blocked skills?
It refuses to install them and requires explicit confirmation before executing. If a skill is blocked, the recommendation says so.
How is OpenAgentSkill different from a list of skills?
It ranks skills for a specific task and agent, shows a score and an install receipt, and takes real run outcomes into account. The author describes it as a decision layer between finding a skill and running it.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI and Python SDK to search, install, create and evaluate reusable agent skills from a shared library
A tool that builds tested skills from a real workflow and governs their lifecycle through your own marketplace
A skill with daily rankings of agent skills by installs, growth and mentions across several platforms at once
A CLI installer and package manager for coding agent skills: one command installs skills into many runtimes and helps build your own
A CLI skill manager for AI agents: install from repositories, translate between agent formats and deploy into their configs
A Go terminal UI for agent skills: list, check, update and find new ones in a single window
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| OpenAgentSkill (this entry) | no | 366 | ||
| OpenSkills | no | 10,813 | installs skills from GitHub repositories but does no matching or trust scoring | |
| SkillKit | no | 1,566 | a skill manager for many agents, without service-side task matching | |
| HeyClaude MCP | no | not required | 304 | a read-only MCP server for the HeyClaude registry |
README in short
The author opens with the thesis that choosing a skill is harder than publishing one and shows a demo API request. The site serves separate machine-readable endpoints, and developers get a TypeScript SDK.