uinaf/verify
Run the builder-owned pre-review verification pass for a completed change using repo guardrails and real-surface evidence. Use a separate evaluator for complex, subjective, or high-risk work; when the repo is not verifiable, report blocked with the exact missing infrastructure and required setup.
92
—
Does it follow best practices?
Impact
92%
1.02xAverage score across 3 eval scenarios
Passed
No findings from the security scan
Evaluation results
89%
-6%Blocked Verification
Blocked verdict
100%
100%
Commands attempted
100%
100%
Missing infra evidence
100%
100%
No static proof substitution
85%
92%
Readiness gaps listed
100%
100%
Infrastructure next steps
100%
100%
Surfaces not exercised honestly
100%
90%
No invented tests
100%
50%
Compact blocked footer
62%
50%
92%
6%Calculator Library Verification
Guardrails run first
86%
100%
Test output recorded
100%
100%
Failure path exercised
100%
100%
Verdict field present
100%
100%
Change Verified section
100%
100%
Surfaces Exercised section
100%
100%
Self-Corrections section
50%
62%
Exact Evidence section
87%
100%
Recommended Follow-up section
100%
100%
Compact verification footer
62%
50%
Error actionability assessed
37%
87%
No unverified claims
100%
100%
95%
6%User API Verification
Server started and queried
73%
100%
Real HTTP client used
100%
100%
Actual response captured
100%
100%
Error/failure path exercised
100%
100%
Exact surfaces named
100%
100%
Exact commands recorded
100%
100%
Success kept terse
50%
75%
Finding tied to impact
100%
100%
Valid verdict
100%
100%
Compact verification footer
62%
62%
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6