agent-skills-discovery-rfc

Agent Skills Discovery via Well-Known URIs defines a mechanism for discovering agent skills using the `.well-known/agent-skills/` URI path, following RFC 8615. Organizations publish a machine-readable index at a predictable location, eliminating the need to search GitHub repositories, documentation sites, and other scattered sources. The specification supports both individual skill files and archives with supporting resources, using progressive disclosure to manage context efficiently. Agents and tools can discover, fetch, and integrate skills without prior configuration. This standardized approach benefits developers who need consistent skill access and operations teams managing skill integration across AI agents.

[{"step":"Identify the well-known URI endpoint for your agent skills registry. This is typically provided by your agent framework or platform (e.g., `https://api.example.com/.well-known/agent-skills`).","tip":"Check your agent framework's documentation or ask your operations team for the correct URI. If unsure, try common paths like `/.well-known/agent-skills` or `/skills/registry`."},{"step":"Use the prompt template to generate a discovery query. Replace [URI_ENDPOINT] with your actual URI and [TOOL_NAME_OR_SYSTEM] with the tool or system you plan to integrate with (e.g., 'Claude Code', 'GitHub Actions', or 'Kubernetes Operator').","tip":"If your agent framework supports authentication, include the necessary headers or tokens in the request. For example, use `curl -H \"Authorization: Bearer <TOKEN>\" <URI_ENDPOINT>` to test the endpoint manually."},{"step":"Execute the query in your preferred AI assistant or automation tool. For example, paste the prompt into Claude or ChatGPT and replace the placeholders.","tip":"If the endpoint requires specific headers or authentication, ask the AI assistant to include them in the request. For instance: 'Make an HTTP GET request to [URI_ENDPOINT] with headers {Authorization: Bearer <TOKEN>} and return the skills in JSON format.'"},{"step":"Validate the output by checking the skill names, descriptions, and input/output formats. Ensure the skills match your expected use cases and that the output format is compatible with your integration tool.","tip":"Compare the discovered skills against your requirements. If a required skill is missing, check if the URI endpoint is correct or if the skill is registered under a different path."},{"step":"Integrate the discovered skills into your workflow. For example, use the skills in an automation pipeline, a CI/CD process, or a custom agent application.","tip":"Store the skills metadata in a configuration file or database for easy reference. For example, save the JSON output to `agent-skills.json` and reference it in your automation scripts."}]