VibeCop runs five LLM-powered detector agents against every pull request: hallucinated abstractions, semantic duplication, over-engineering, pattern drift, and reinvented utilities. Each agent is calibrated to your codebase's Stack Profile, and every finding it produces carries an entropy weight that feeds the Architecture Integrity Index.
Hallucinated abstractions flags layers and wrappers that add indirection without earning it. Semantic duplication flags logic re-implemented instead of reused. Over-engineering flags premature generalization and unused flexibility. Pattern drift flags the same problem solved inconsistently. Reinvented utilities flags hand-rolled code duplicating an existing library or internal helper.
| Agent | Catches |
|---|---|
| Hallucinated Abstractions | Layers, wrappers, interfaces adding indirection without value |
| Semantic Duplication | Logic re-implemented in parallel paths instead of shared |
| Over-Engineering | Premature generalization, unused flexibility, config-for-config's-sake |
| Pattern Drift | Same problem solved differently across the codebase |
| Reinvented Utilities | Hand-rolled logic duplicating an existing library or internal utility |
Four deterministic hygiene layers run first, fast and free of false positives from hallucinated rules: dependency CVE scanning against the live osv.dev database with EPSS and CISA KEV enrichment, secret detection, static application security testing (SAST) across 18 CWE- and OWASP-tagged rules, and infrastructure-as-code misconfiguration checks.
No — hygiene findings are tracked in a separate tray from architectural findings. They matter for security and compliance, but only findings from the five architectural detector agents carry an entropy weight that affects the index; see the Architecture Integrity Index doc for the full scoring formula.