review-process

To deduplicate code review findings from multiple AI agents, this Skill applies deterministic fingerprinting and occurrence-key generation to reconcile findings that drift in wording across cycles. It anchors findings to code using locator-based matching and enforces semantic deduplication rules to produce a concise reconciliation set.

A deterministic fingerprinting system for multi-model code review assigns stable identity markers (fp_v1) and occurrence keys (occ_v1) to findings raised by automated reviewers. It ensures collision-safe suppression and cross-session tracking by applying robust normalization, evidence verification, and locator anchoring.

To handle conflicting evidence from automated security audit reviewers, this Skill verifies evidence and can auto-dismiss findings with unverifiable evidence. It supports catch-up review scenarios to maintain momentum in long-running debates by enforcing suppression and collision handling rules.

Yes, you can use this reconciliation process for long-running adversarial review debates. It supports catch-up review scenarios and cross-session tracking via occurrence keys to maintain debate momentum, applying auto-dismiss for unverifiable evidence and enforcing semantic deduplication to suppress drift.

Automated reviewer findings drift in wording across review cycles due to non-deterministic generation in multi-agent code review contexts. This Skill solves the drift problem by applying exact fingerprinting, locator anchoring, and semantic deduplication rules to reconcile findings across cycles and sessions.

No specific dependencies are required to run the multi-model review reconciliation. This Skill operates as a standalone process module with no external dependencies, enforcing fingerprinting, evidence verification, and deduplication rules directly within the adversarial-review workflow.