Outcome
Practitioners can deploy and operate AURA in production without unresolved blockers in safety, correctness, or core reliability.
Scope
- Human approval for destructive tool calls.
- Production-blocking defects that corrupt results, report false success, lose required state, or make supported workflows unsafe or unusable.
- Completed capabilities that materially improve repeatable production RCA, including Agent Skills.
Entry criteria
Work belongs in this initiative only when it:
- Blocks a supported production workflow or creates unacceptable operational risk.
- Has a reproducible failure mode or a specific safety contract.
- Can be verified through an integration or regression test.
General engineering debt, benchmark quality, and speculative capability work do not belong here.
Success
- Destructive actions fail closed unless an authorized human approves them.
- Production-blocking defects are fixed and protected by regression coverage.
- Supported production workflows complete without known critical blockers.
Outcome
Practitioners can deploy and operate AURA in production without unresolved blockers in safety, correctness, or core reliability.
Scope
Entry criteria
Work belongs in this initiative only when it:
General engineering debt, benchmark quality, and speculative capability work do not belong here.
Success