uinaf/autoreview
Run the bundled Codex/Claude autoreview helper as a structured second-model closeout. Use for explicit autoreview or tool-backed final-review requests after builder proof exists; it validates findings and reruns focused tests but does not issue a ship decision.
98
98%
Does it follow best practices?
Impact
98%
1.44xAverage score across 4 eval scenarios
Passed
No known issues
Evaluation results
100%
15%Autoreview Finding Triage
Accepted in-scope blocker
100%
100%
Architecture churn rejected
100%
100%
Sibling risk follow-up
92%
100%
Scope baseline referenced
100%
100%
Verification after fix
100%
100%
Autoreview rerun
0%
100%
No public posting
100%
100%
Concise final report
75%
100%
100%
85%Autoreview Setup and Target Selection
Autoreview mode and engine selection
AUTOREVIEW variable set
0%
100%
AUTOREVIEW_HARNESS variable set
0%
100%
AGENTS_HOME for global skill
0%
100%
$AUTOREVIEW invocation used
0%
100%
Local mode for uncommitted changes
0%
100%
Branch mode for PR/feature branch
0%
100%
Branch mode has --base flag
0%
100%
Commit mode for pushed-to-main
0%
100%
--commit HEAD for commit mode
0%
100%
No --engine claude override
100%
100%
No panel flags
100%
100%
92%
7%Autoreview Target Selection
Branch target selected
77%
100%
PR base discovery command
100%
100%
Autoreview helper used
78%
78%
No empty local review
100%
100%
No push requirement
100%
100%
Focused proof loop
78%
100%
Reject advisory weakness
60%
50%
Compact closeout shape
87%
100%
100%
10%Release Branch Autoreview Closeout
Release freeze recognized
100%
100%
Crash finding accepted
100%
100%
Style finding deferred
100%
100%
Smallest allowed fix
100%
100%
Focused release proof
100%
100%
Autoreview rerun after fix
0%
100%
Forward-port decision
100%
100%
No branch broadening
100%
100%
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6