evals-testing / pattern
prove-it-side-effect-check
Capability: prove-it-side-effect-check
Use it when
- require agents to prove side effects before claiming done
- catch agent-said-done-not-done with evidence checks
What it solves
Not the fit when
- replaces product QA
Install
n/a, checklist + pair with promptfoo/deepeval
Invoke
1) Agent claims done 2) Re-read CRM/DB/email/outbox 3) Assert side effect in tests 4) Only then close ticket
Alternatives
No reviewed alternatives recorded yet.