01 / DECIDE
Architecture decision
Use it when prompts are shared across users, languages, tools, or releases.
Version durable behavior, delimit untrusted input, and specify an acceptance format so prompt changes remain reviewable. Map input, output, state, tools, and side effects as one observable path, then name the owner of separate instructions from user data.
02 / OPERATE
Three-step field procedure
- 01
Lock the contract
Write down the caller, data classification, success condition, timeout, cancellation, and ownership. Use it when prompts are shared across users, languages, tools, or releases.
- 02
Run one bounded path
Build one end-to-end path with correlation, typed state, and reversible failure handling. Version durable behavior, delimit untrusted input, and specify an acceptance format so prompt changes remain reviewable.
- 03
Verify the signal
Turn acceptance into a repeatable fixture, contract test, or browser test. A prompt suite compares task outcomes, injection resistance, and refusal handling before release.
03 / BOUND
Operational boundary
Instructions are not a security boundary; server code still owns authorization and side effects.
04 / PROVE
Acceptance evidence
A prompt suite compares task outcomes, injection resistance, and refusal handling before release.
SOURCE / HTTP
Reproducible source probe
curl -fsSI 'https://docs.x.ai/developers/model-capabilities/text/generate-text' | sed -n '1,5p'