Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, but long-horizon tasks require more than action prediction. An agent must coordinate perception, planning, execution, progress verification, and recovery as the physical state evolves. A new framework called EmbodiedSkills addresses this by treating each skill decision as an execution proposal: the runtime checks prerequisites before execution and verifies the outcome afterward.
EmbodiedSkills connects high-level skill selection, bounded low-level VLA execution, and post-action verification through a shared executable-skill interface. Because this interface remains fixed, low-level VLA policies can be replaced or adapted without changing the agent loop. The interface also records planning, execution, verification, and recovery events as structured trajectories, which provide supervision for individual components and support optional online adaptation when interactive feedback is available.
The framework operates in a six-stage loop: Observe, Plan, Preflight, Execute, Verify, and Recover. The planner breaks an instruction into semantic subtasks, and the VLA executes actions conditioned on the current subtask. Before execution, the system checks readiness; after execution, fresh observations support progress verification. An unfinished subtask can continue with another action chunk, a completed subtask advances the plan, and an unsuccessful attempt can trigger recovery or replanning.
The authors instantiate EmbodiedSkills with Qwen3-VL and OpenPI/pi0.5 on RoboTwin 2.0 and LIBERO. Task-adapted low-level VLA policies achieve an average success rate of 86.20% across 50 RoboTwin 2.0 tasks and 97.40% across the four LIBERO suites. On four memory-dependent RMBench tasks, the same approach achieves 12.5% average success, highlighting the need for the framework's verification and recovery mechanisms.
The same interface connects execution with training: subtask-aligned demonstrations train the low-level action policy, while structured interaction trajectories provide supervision for high-level skill selection. The scheduler can be fine-tuned with the action backend frozen, allowing independent adaptation of planning, verification, and action components. This makes EmbodiedSkills a trainable and inspectable agent layer for turning VLA policies into closed-loop embodied systems.