01 / DECIDE
Architecture decision
Use it for parallel research, test analysis, review, and independent components with no write collision.
Subagents need independent ownership and result contracts; headless CI needs fixed permissions, machine-readable output, and isolated credentials. Map input, output, state, and side effects as one observable path before deciding which layer owns parallelize evidence, not conflicting writes.
02 / BUILD
Three-step implementation
- 01
Frame the contract
Write down the caller, data classification, success condition, timeout, cancellation, and ownership. Use it for parallel research, test analysis, review, and independent components with no write collision.
- 02
Build one narrow path
Implement one end-to-end path with request correlation, typed state, and reversible failure handling. Subagents need independent ownership and result contracts; headless CI needs fixed permissions, machine-readable output, and isolated credentials.
- 03
Prove the outcome
Turn acceptance into a repeatable fixture, contract test, or browser test. CI artifacts show task ownership, structured outcomes, tests, diffs, and that every worker stayed inside its boundary.
03 / BOUND
Production boundary
Never expose untrusted pull requests to privileged secrets or let multiple workers modify the same files without coordination.
04 / PROVE
Acceptance evidence
CI artifacts show task ownership, structured outcomes, tests, diffs, and that every worker stayed inside its boundary.
SOURCE / HTTP
Reproducible source probe
curl -fsSI 'https://learn.chatgpt.com/docs/agent-configuration/subagents' | sed -n '1,5p'