01Result
Three-phase pipeline: detect → analyze → brief
AI Code Review · Go · Python · TS
The missing code review layer — AI that reviews what the AI wrote. Detects AI-generated code in PRs, analyzes context blindness, and briefs human reviewers.
License
MIT
Detect
~2ms / file (Go)
Analyze
~5s high-risk section
Runtime
Docker Compose
Interactive lens
Drag the handle right to reveal the solution over the problem.
Problem
AI tools write syntactically impressive code that is often semantically dangerous — migrations that ignore table scale, retries against non-idempotent APIs, config flips with wide blast radius. Linters catch syntax; they miss operational context. Human senior review does not scale with Copilot volume.
Solution
TRIBUNAL acts as an automated senior engineer in CI: DETECT AI authorship, ANALYZE context blindness against topology/incidents/runbooks, BRIEF the human reviewer so review time drops from roughly 30 minutes to about 30 seconds per change with full operational context.
Drag right to reveal the solution · 12%
Three-phase pipeline: detect → analyze → brief
Flags scale, idempotency, cascade, incident-pattern, and race-condition blindness
Posts human-readable briefings into the PR review flow
Execution flow
End-to-end flow from trigger to decision output.
Step 1 of 6
PR event from GitHub / GitLab / Gitea
Step 2 of 6
Go interceptor: webhook, changed files, 3-signal AI authorship detection
Step 3 of 6
Go context graph: topology, incidents, dependencies
Step 4 of 6
PostgreSQL context layer for persistent correlations
Step 5 of 6
Python semantic analyzer (Claude) + briefing generator
Step 6 of 6
OutputTypeScript PR overlay: check runs, annotations, risk heatmap
Capabilities
Scale blindness — e.g. migrations on multi-billion-row tables with zero-downtime constraints
Idempotency blindness — retries against non-idempotent payment/API calls
Cascade blindness — config/flag changes spanning dozens of services
Incident pattern blindness — repeating known failure modes
Dependency and race-condition blindness in shared libraries and concurrent code
Pin a layer
Click a layer to isolate it — MIT · Go · Python · TypeScript
Live signal
1 TRIBUNAL fills the gap between perfect syntax and catastrophic semantics. It specializes in what AI-generated diffs did not know — service scale, incident history, dependency blast radius — and briefs humans so review capacity can keep up with AI authorship.
[ VERIFY ]
Claims should survive a read of the repo.