from Claude Code Skills444
Merges and deduplicates evaluation worker results into a single verified review summary for coordinators, applying an architecture gate and verification checks.
This skill processes and merges outputs from multiple evaluation workers (e.g., ln-311..ln-314), deduplicates overlapping findings against prior review history, verifies agent suggestions using a Critical Verification protocol, and emits a single verified review-merge summary for downstream coordinators. It enforces an Architecture Gate to ensure accepted suggestions implement correct architecture without legacy shims.
Use this worker during multi-agent evaluation pipelines when multiple evidence lanes or worker summaries finish and a single authoritative, deduplicated review must be produced. Particularly useful in automated review systems, CI analysis pipelines, and multi-model research runs where verification and deduplication are required before publishing results.
evaluation_worker_runtime_contract.md, evaluation_summary_contract.md, agent_delegation_pattern.md).identifier, phase_order, summary_kind=review-merge, operation=merge). The skill documents payload fields to emit and checks to perform before marking pass=true.Likely used by Code-oriented orchestrators and agent runtimes (Codex/Copilot-style automation, Claude Code orchestrations) integrated with evaluation-worker-runtime families.
This skill has not been reviewed by our automated audit pipeline yet.
