01 / DECIDE
Architecture decision
Use it for calculations, data transformation, charts, and file inspection where deterministic code improves the answer.
Send bounded inputs, define expected artifacts, and validate outputs outside the sandbox before they enter business workflows. Map input, output, state, and side effects as one observable path before deciding which layer owns use code execution as a disposable compute cell.
02 / BUILD
Three-step implementation
- 01
Frame the contract
Write down the caller, data classification, success condition, timeout, cancellation, and ownership. Use it for calculations, data transformation, charts, and file inspection where deterministic code improves the answer.
- 02
Build one narrow path
Implement one end-to-end path with request correlation, typed state, and reversible failure handling. Send bounded inputs, define expected artifacts, and validate outputs outside the sandbox before they enter business workflows.
- 03
Prove the outcome
Turn acceptance into a repeatable fixture, contract test, or browser test. Fixtures verify calculations independently and reject missing, oversized, malformed, or unexpected artifacts.
03 / BOUND
Production boundary
A sandbox does not make unknown code trustworthy; restrict inputs, output types, size, runtime, and downstream effects.
04 / PROVE
Acceptance evidence
Fixtures verify calculations independently and reject missing, oversized, malformed, or unexpected artifacts.
SOURCE / HTTP
Reproducible source probe
curl -fsSI 'https://developers.openai.com/api/docs/guides/tools' | sed -n '1,5p'