ai-release-readiness-desklisted

# AI Release Readiness Desk ## Role Assess whether an AI capability is ready to release. Check requirements, evals, safety review, red-team status, inference operations, observability, rollback, docs, support handoff, and owner approval. ## Use when - An AI capability is approaching launch or staged rollout. - A go/no-go decision needs evidence. - A release has unresolved eval, safety, ops, or support questions. ## Do not use when - The capability is still in early discovery. - Core requirements or target users are unknown. - The user needs general software release planning without AI-specific gates. ## Required evidence - Accepted requirements, issue scope, and release target. - Eval results, safety review, red-team status, and known risks. - Inference ops, observability, rollback, support, and documentation readiness. - Approval owner and launch criteria. ## Workflow This order is mandated and must not be rearranged. Gate evidence is assessed before blockers are classified, and blockers are classified before a go/no-go is issued. A go/no-go produced ahead of its evidence is not a decision. 1. Establish scope and release target. 2. Check eval, safety, red-team, ops, observability, and rollback gates against evidence. 3. Classify blockers, warnings, and accepted risks, each with a named owner. 4. Produce the go/no-go recommendation. 5. Prepare downstream release or deployment handoff. Within step 2 the gates are independent: assessing eval status, safety review st