What problem does it solve? Automated trading systems fail in expensive, non-obvious ways: mock data leaking into production paths, AI output trusted without validation, race conditions doubling exposure, duplicate order submissions, and risk gates that can be bypassed. Generic code review instincts miss these domain-specific failure modes, so working on the Viipers multi-agent trading terminal requires a structured audit methodology and enforced safety invariants. ## Core Features & Use Cases - Full production-readiness audits: Builds and maintains a living checklist at docs/TRADING_SYSTEM_AUDIT.md covering agents, risk engine, database concurrency, exchange integration, security, and testing, with evidence-based marking conventions. - Risk-gate invariant enforcement: Ensures AI agents can only propose trades while a deterministic risk engine approves them and only the order-executor agent reaches the broker, flagging any bypass path as critical. - Narrow task support: Applies the same core principles (idempotency, LLM output validation, fail-loud error handling) to smaller changes like adding an agent, fixing a duplicate-order bug, or reviewing a PR. - Use Case: When asked to make the Viipers repo production-ready, the skill guides a stage-by-stage trace of the consensus pipeline (sentiment, analysis, coordination, risk, execution), hunts for hidden mocks and race conditions, and closes out with an executive summary of verified and unverifiable items. ## Quick Start Ask the AI to run a full production-readiness audit of the Viipers trading terminal and create the audit checklist document.
