Agent Prompt & Contract

Pin down the literal source code of an AI agent — the system prompt, the tool/function schemas, and the structured input/output contracts. Use whenever the user is ready to turn an architecture into something an engineer can type against: writing or refining the agent's system prompt, defining the JSON schemas for each tool the agent calls, specifying the request/response contract, or locking the structured-output format. Trigger on "system prompt", "tool schema", "function calling", "I/O contract", "output schema", "what does the agent actually receive and return", even without those exact words. This is the code-ready keystone of the agent-dev-framework pipeline — slot …