Skills

Reusable capability bundles with explicit execution boundaries

Skills let you attach reusable capabilities to inference requests without resending large tool payloads every time. Each skill has a clear execution boundary so you always know what runs inside TheRouter and what runs in your infrastructure.

Skill types

  • Built-in Skills (managed by TheRouter): Capabilities like web search, file parsing, and response healing that run inside TheRouter infrastructure. Enable and configure — no custom code.
  • Schema Bundles: Reusable sets of tool definitions and optional instructions that get injected into requests. No server-side execution — the model decides when to use the tools.
  • Remote Webhooks (coming soon): Executable skills hosted in your own infrastructure. TheRouter sends signed HTTPS requests to your endpoints during managed execution loops.
skills-request.json
{
  "model": "anthropic/claude-sonnet-4.5",
  "messages": [{ "role": "user", "content": "Summarize current AI regulation changes." }],
  "skills": [{ "id": "web" }, { "id": "response-healing" }]
}

Built-in skills

  • Web Search: Attach fresh web sources and citations to model responses.
  • File Parser: Parse PDFs and files into model-ready text chunks.
  • Response Healing: Repair malformed JSON before returning output to your application.
TypeScript
const completion = await client.chat.completions.create({
  model: "anthropic/claude-sonnet-4.5",
  messages: [{ role: "user", content: "Extract key actions from this PDF." }],
  // Skills is a TheRouter extension on /v1/chat/completions and /v1/responses
  skills: [{ id: "file-parser" }],
});
TheRouter extension
The skills field is a TheRouter extension on /v1/chat/completions and /v1/responses. Standard OpenAI-compatible fields remain unchanged. Requests without skills behave exactly as before.

Deep guide

Ready to go deeper? Follow the skills learning path: