What problem does it solve? High-stakes engineering changes—core engine modifications, breaking API contracts, cross-workspace refactors—often ship without rigorous pre-validation against reliability, scalability, security, observability, and maintainability, leading to production incidents and costly rework. ## Core Features & Use Cases - PGE Iterative Refinement: A Planner-Generator-Evaluator loop produces and critiques architectural proposals against a Sprint Contract with explicit acceptance criteria, capped at 3 iterations. - Council Mode with Harness Delegates: Three isolated delegates (Alpha for Reliability, Beta for Scalability, Gamma for Security/Observability) submit proposals, conduct anonymous peer review, debate disagreements, and reach ACK-ALL consensus with veto and override rules. - Harness Verification Gate: An explicit five-pillar checklist (reliability, scalability, security, observability, maintainability) must pass before final output, with transcripts persisted to .git/.orchestration/harness/. - Token Ring Mode: For ambiguous requirements, a token circulates among delegates until a full cycle produces no changes, signaling emergent consensus. - Use Case: When asked to "harness the best approach for real-time state sync," the skill runs the pge-council pipeline: Alpha proposes a WebSocket/polling hybrid, Beta challenges concurrency at scale, Gamma mandates per-room trace logging, and the gate confirms all five pillars pass before delivery. ## Quick Start Ask the assistant to run a harness deliberation on your proposed architecture for a core engine change, such as "harness this new matching algorithm design before I implement it."
