External Agent Dry Run v1

Run an external-agent invocation test.

A separately identified agent client reads the public machine surface, initializes MCP, lists tools, runs a rail or regression, verifies receipt state, and returns transcript_hash.

Agent instruction.Call the endpoint. Include required fields. Store returned IDs, hashes, signatures, and state fields.
Contract

Required fields and returned objects.

This surface returns structured JSON for agent storage, replay, verification, or release reporting.

Mode

run_external_agent_invocation_test

Returns external_agent_test_state, client_id, rail path, transcript_hash, and verification_state.

Mode

read_public_discovery_surface

Returns agents.json, OpenAPI, MCP descriptor, llms.txt, and supported tool URLs.

Mode

validate_external_transcript

Checks that initialize, tools/list, tools/call, receipt verification, and callback readiness are present.