KAIROS development — agent instructions for kairos-dev MCP

Use this skill when an AI agent should run protocols or workflow tests against kairos-dev (the KAIROS MCP in this repo’s dev environment). All instructions assume MCP-only usage: no shell, no filesystem except writing to reports/.

When to use which flow

| Flow | File | Use when | |------|------|----------| | AI–MCP integration | ai-mcp-integration.md | Agent runs protocols one by one from a real-user perspective: import each protocol from docs/examples/, run it to completion, write one report per protocol under reports/<run-id>/<protocol-folder>/report.md with inline pretty-printed JSON. Tighter scope (no README, one report per protocol). | | Workflow test | workflow-test.md | Agent runs the four scenarios (imports; search + workflows; update step; update chain). Single summary report at reports/<run-id>/report.md plus per-call JSON under reports/<run-id>/calls/. Use for dev/qa workflow validation. |

• For one-by-one protocol runs (e.g. scripted or “run each example as a user”): use ai-mcp-integration.md.
• For full workflow test (four scenarios, one run): use workflow-test.md.

Related

• Mintable protocol examples: docs/examples/protocol-example-*.md (do not read docs/examples/README.md when under ai-mcp-integration constraints).
• Test harness and automated import test: tests/workflow-test/ and npm run dev:ai-mcp-integration / npm run dev:test -- tests/integration/kairos-mint-docs-examples.test.ts.
• MCP payload (tools, prompts, resources) is in src/embed-docs/ and served by the server at runtime.