Venue Evidence Audit
Run only when the Orchestrator authorizes the audit and P0 is substantially complete. Start with python tools/context_view.py show --role venue, then read referenced Claim-Evidence and gap artifacts. Read docs/PROTOCOLS.md, docs/ROLE_HANDOFFS.md, docs/CONTEXT_POLICY.md and docs/EVIDENCE_COMPLETION_PROTOCOL.md on a fresh context or changed hash. Reuse an existing dated source set only when its task, venue, freshness and coverage still fit; search further for changed expectations or unresolved gaps. Context savings never justify skipping relevant papers or negative evidence.
Analyze 5-10 recent, highly relevant papers, prioritizing same task, research question, method class, venue and recency. Record search date, sources, relevance, inclusion/exclusion reasons and uncertainty. Stop once core expectations, material gaps and high-ROI actions are stable; do not expand into a systematic review. Compare datasets, baselines, metrics, seed/statistical practice, ablations, sensitivity, robustness, generalization, efficiency, complexity, qualitative analysis and figure/table conventions.
Produce an Evidence Gap Matrix with current status, venue frequency, Claim relevance, reviewer risk, estimated cost, information gain and recommended priority. Recommendations do not enter the queue automatically. A common but Claim-irrelevant experiment is P3/SKIP. Do not alter the Research Spine, Primary Claims or Method, and report any positioning concern only as risk.
