Get Me a Skill
Find and assess agent skills for any job. The skill clarifies what the work requires, searches existing skills, inspects likely matches, and recommends whether to use, combine, adapt, or build. It treats skills.sh as a discovery source, not a quality verdict. This project is independent of skills.sh and Vercel's find-skills skill.
The canonical workflow is the open-standard SKILL.md. The plugin.json and marketplace entry package that same skill for ChatGPT and Codex; they are not required by other agents. No hosted service, API key, or agent-specific script is bundled.
Install across agents
Use a global installation when you want the skill in all projects. The Skills CLI accepts agent targets; select the agent you use instead of installing to every agent automatically.
| Host | Global installation or skill location | Invocation |
|---|---|---|
| Claude Code | npx skills add developer-mohanaditya/get-me-a-skill -g -a claude-code -s get-me-a-skill | Select or invoke the skill by name. |
| Codex, OpenCode, Freebuff CLI, DeepSeek Harness, Deep Code | npx skills add developer-mohanaditya/get-me-a-skill -g -a universal -s get-me-a-skill installs in the shared ~/.agents/skills/ directory these hosts can read. | Codex: $get-me-a-skill; OpenCode: select the skill; Freebuff: /skill:get-me-a-skill; Deep Code: select from /. |
| Google Antigravity app | Copy skills/get-me-a-skill/ to ~/.gemini/config/skills/get-me-a-skill/. Project installation in .agents/skills/ also works. | Select the skill in Antigravity. |
| Other Agent Skills-compatible hosts | Use npx skills add developer-mohanaditya/get-me-a-skill -g -a <agent-id> -s get-me-a-skill, or copy the skill folder to the host's documented skill directory. | Use the host's skill picker or explicit invocation syntax. |
DeepSeek is also a model provider. A model does not install skills by itself; the host running it must load Agent Skills. Deep Code and DeepSeek Harness have compatible skill directories. For an agent without native skill loading, provide the SKILL.md text as instructions for a single conversation; automatic discovery and reuse will not be available.
For the ChatGPT/Codex plugin package, add this repository as a marketplace and install the plugin:
codex plugin marketplace add developer-mohanaditya/get-me-a-skill --ref main
codex plugin add get-me-a-skill@get-me-a-skill
In ChatGPT, select @get-me-a-skill when the plugin is enabled. Include the job, required depth, and constraints. The skill may ask a few focused questions if they would change the match.
Quality across hosts
The same instructions and hard gates apply everywhere. A full assessment requires current search, access to complete candidate skill files, and enough tools to verify claims or run a bounded trial. If a host lacks these capabilities, the skill must disclose the gap and produce a limited assessment. Installing the file cannot make different models or tools produce identical results.
Use evals/cases.md to compare hosts. Test both discovery and the final recommendation; seeing the skill in a menu proves installation only.
Maintain it
Edit the canonical skill in skills/get-me-a-skill/, validate it against the Agent Skills specification, and rerun the evaluation cases in each supported host before claiming parity. Bump plugin.json's version for behavior changes. Users of the Skills CLI can run npx skills update get-me-a-skill -g; Codex plugin users can refresh the marketplace with codex plugin marketplace upgrade get-me-a-skill and update the installed plugin. Keep host paths current as agents change their discovery rules.
License
MIT. See LICENSE.