Good task boundaries end in evidence: a map, a decision, a working change or a verified result.
The decision behind the technique
One giant request accumulates irrelevant context and makes failures expensive. Dozens of tiny prompts create coordination overhead and lose the original acceptance criteria.
Good task boundaries end in evidence: a map, a decision, a working change or a verified result.
A practical step-by-step workflow
Use this sequence as a default, then tighten or expand it only when the evidence changes.
- Write one outcome and its acceptance tests
- Separate exploration, implementation and verification
- Give each stage a concrete deliverable
- Carry forward only decisions and evidence needed by the next stage
Failure boundary and finish line
Do not split a tightly coupled migration across agents that will edit the same state concurrently.
Finish when the next action, owner, evidence and stop condition are all explicit.
Frequently asked questions
What should I do first?+
Write one outcome and its acceptance tests
What is the main failure mode?+
Do not split a tightly coupled migration across agents that will edit the same state concurrently.
OFFICIAL SOURCES
Verify against the source
OpenAI Codex best practicesThis guide provides general information, not an OpenAI allowance promise or support decision. Your account and official pages govern actual usage, benefits and recovery times.