What problem does it solve? Managing eval-gated experiments in Caveman Cloud requires reading evidence before any lifecycle change, but the current control API does not yet enforce transition and evidence gates atomically, so unsafe mutations must be blocked. ## Core Features & Use Cases - Evidence-first inspection: Load project context, experiment state, and results via MCP tools or the caveman CLI before any recommendation. - Gated recommendations: Propose exactly one action (start, approve, cancel, or rollback) only when evidence is complete and passing; report "evidence incomplete" otherwise. - Mutation blocking: Refuse to execute lifecycle commands because agent-generatable approval strings are not proof of human intent and server-side gates are not yet enforced. - Use Case: A user asks whether experiment 7f... can be promoted. The Skill reads its state and results, checks guardrails and quality evidence, then reports a supported recommendation while blocking execution. ## Quick Start Ask the assistant to review Caveman experiment 7f and recommend whether it should be approved based on its current evidence.
