rlxp-make-contract

An RL experiment contract formalizes task_id, study_id, metrics, guardrails, baseline, evaluation protocol, budget, hardware, and stopping rules into a bounded scope. It prevents training or autonomous execution from starting until every required approval and constraint is explicitly recorded.

Study scoping for audit handoff turns an approved task definition and constraint set into a bounded contract. It establishes repo-local state for candidates, runs, audits, and reports, ensuring baseline agreement and metric scoping remain explicit, comparable, and safe.

Training approval control requires explicit budget, hardware, evaluation, and automation fields in the experiment contract. Safety gating blocks any training or autonomous execution until all required approvals and constraints are recorded in the study scope.

You can update an existing RL experiment contract when a metric change requires a fresh study scope. The contract creation process applies to both new task definitions and updates to existing studies, re-evaluating all required fields before enabling training.

A confirmed task definition and constraint set are required before initializing RL study planning. The contract creation process needs explicit task, metric, guardrail, budget, hardware, baseline, evaluation, and automation fields to proceed with study initialization.

Training fails to start when safety gating detects missing required fields. The contract must contain explicit task, metric, guardrail, budget, hardware, baseline, evaluation, and automation entries before training or autonomous execution can be enabled.