rlxp-audit-task

To audit a reinforcement learning repository, inspect entrypoints, configs, reward logic, termination conditions, and curriculum to infer the actual task. This captures evidence needed for a scoped experiment setup without launching training.

A task card in reinforcement learning is a scoped documentation artifact that aligns a user objective with an implemented environment, reward, and evaluation setup. You need one when onboarding an unfamiliar repository or defining a new RL study.

Yes, you can generate an adapter draft from an RL codebase without running training jobs. The audit inspects logging, checkpoints, and metrics to produce a task-specific adapter draft and task card for contract-based experiment planning.

To map an unclear reward structure in a reinforcement learning repository, inspect the reward logic and domain randomization configurations within the codebase. This audit clarifies the evaluation protocol before any optimization begins.

Yes, repository audit supports new reinforcement learning objectives and curriculum implementations by capturing open questions and preparing the project for a fresh task_id. It ensures objectives do not mix by producing scoped documentation.

A limitation of auditing an RL repository for task discovery is that it generates only a task card, adapter draft, and open questions without launching training jobs. It prepares the project for experiment planning but does not execute optimization.