rl-experiment-loop

Reinforcement learning experiment loop coordination organizes audits, candidate proposals, validation, monitoring, and reporting across a single lifecycle. It enforces contract approval, budget limits, and metric invariants before any launch-capable transition.

Unsafe reinforcement learning run launches are prevented by enforcing contract approval, budget limits, held-out metric invariants, and an audited runner boundary before allowing any launch-capable transition.

Monitoring and auditing reinforcement learning candidates inside task directories requires organizing candidate validation, run monitoring, and result auditing within .rlxp task directories to preserve the lifecycle of approved candidates.

Older reinforcement learning prompts work with a structured experiment loop coordinator through legacy compatibility, which keeps existing prompts functioning while steering new work toward the newer rlxp-autoloop path.

A task- and study-scoped reinforcement learning experiment loop is needed when an approved RL study requires a monitored cycle from draft candidate to validated launch, then back through analysis and reporting.

Automating report updates after reinforcement learning run auditing is best handled by integrating reporting into the experiment loop lifecycle, ensuring reports reflect validated candidates and audited results within .rlxp task directories.