Adaptation to complex tasks and multiple scenarios remains a significant challenge for a single robot agent. The ability to acquire organize, and switch between a wide range of skills in real time, particularly in dynamic environments, has become a fundamental requirement for embodied intelligence. We introduce OpenGo, an OpenClaw-powered embodied robotic dog capable of switching skills in real time according to the scene and task instructions. Specifically, the agent is equipped with (1) a customizable skill library with easy skill import and autonomous skill validation, (2) a dispatcher that selects and invokes different skills according to task prompts or language instructions, and (3) a self-learning framework that fine-tunes skills based on task completion and human feedback. We deploy the agent in Unitree's Go2 robotic dog and validate its capabilities in self-checking and switching of skills autonomously. In addition, by integrating Feishu-platform communication, we enable natural-language guidance and human feedback, allowing inexperienced users to control the robotic dog through simple instructions.
Figures & tables
Figure 1: Demonstration of LLM-driven skill execution and composition on the quadruped robot. Given natural language instructions (left), the system interprets user order and transmit it to Unitree Go2. From top to bottom, the commands are: Move Forward, Turn Around, Backflip, and Dance.
Figure 2: Overview of the OpenGo framework. OpenGo is built upon OpenClaw and organized around two core modules, namely the Dispatcher and Memory/State Check . Through a communication platform (e.g., Feishu), human users provide task descriptions, instructions, and execution orders to the system. The Dispatcher selects appropriate skills from the robot-side skill library, while the Memory/State Check module monitors execution status and feeds state information back to support closed-loop decision making. On the robot side, the framework interfaces with perception, controller, state estimation, skill library, and an internal Safety Tool , which serves as an emergency-stop trigger when the robot enters a dangerous state, thereby enabling controllable and robust skill execution on the quadruped platform.
Figure 3: Skill library design in OpenGo. New skills are incorporated through code review and simulation-based validation before entering the skill library. Each skill is organized with structured fields, including skill heads , parameters , constraints , function , and prompts . The parameters are adjustable for task adaptation, whereas the function remains fixed for stable and controllable execution.
Figure 4: Dispatch mechanism of OpenGo. Task descriptions, human instructions, and scene information inferred from perception are jointly provided to the LLM-based dispatcher. The LLM selects appropriate skills from the skill library and organizes them into a step-by-step execution sequence. Execution feedback, including error logs and finished signals, is then returned to the dispatcher, enabling dynamic replanning and closed-loop skill scheduling.
Figure 5: System Latency Analysis of Single-Skill. Experiments are conducted on a real-world Unitree Go2 platform. Each action is executed 10 times, and the response time is measured from the moment the user issues the instruction to the initiation of the robot’s execution.
Figure 6: System Latency Analysis of Multi-Task. We analyze the response time for compositions of 2–4 individual instructions. The order of certain instructions is varied, and each combination is executed five times after the system reaches a stable state.
Transferring robotic skills from simulation to reality requires task knowledge that remains usable across differences in perception, dynamics, and embodiment. We introduce Skill2Real, an agentic policy framework that learns executable skills through a shared application programming interface (API). A Proposer-Verifier-Governor (PVG) loop uses privileged simulation evidence to diagnose outcomes and validate updates, while keeping learned skills grounded in public observations and API semantics. The Cerebellum first acquires local manipulation skills; the Brain then learns task-level composition with the Cerebellum frozen. Both memories transfer to the real robot without task-policy fine-tuning or skill-memory updates. As GPT-5.6 Sol learns skills on LIBERO-90, evaluating each frozen checkpoint with GPT-6 Astra raises LIBERO-Pro Long success from 2.0% to 56.3%, without training on Pro Long. Independent Robosuite training reaches 85.1% and 89.4% mean success with Sol and Opus 5 across seven tasks, respectively. Frozen Sol-trained LIBERO-90 skills achieve 78.75% mean completion across four real-world manipulation tasks with Astra. Removing the Verifier or Governor during LIBERO-90 training lowers final Pro Long success by 17.3 and 13.3 percentage points, respectively. These results support learning and transferring a hierarchy of executable skills through a common robot interface.
Xincheng He, Siyu Ma, Chang Yu +5
University of California, Los Angeles · Shanghai Jiao Tong University · University of Utah +1
Self-evolving agents requires adaptation after deployment, but existing approaches assume a usable learning loop, such as curated skills, successful trajectories, or verifier signals. Real open-world deployments may provide none of these, offering only a task prompt. In this work, we study open-world self-evolution, where an agent must build both its skills and its own verification signals from scratch, using open-world resources but no target-task supervision. We propose OpenSkill, a framework that bootstraps this loop: it acquires grounded knowledge and verification anchors from documentation, repositories, and the web, synthesizes them into transferable skills, and refines those skills against self-built virtual tasks grounded in the anchors rather than in target answers. The open world thus supplies both the knowledge to be learned and a supervision-independent practice environment, with target-task supervision reserved for final evaluation. Across three benchmarks and two target agents, OpenSkill attains the best automated pass rate while satisfying the no-supervision constraint. Analysis shows its skills transfer across models without model-specific adaptation, and its self-built verifier aligns with ground-truth outcomes despite never accessing them.
Zhiling Yan, Dingjie Song, Hanrong Zhang +8
Lehigh University · University of Illinois Chicago · University of British Columbia +3
Vision-language-action and world-action models have demonstrated impressive capabilities in robotics, yet generalization to unseen tasks remains challenging. More recently, general-purpose multimodal agents have shown great potential for zero-shot robotic task solving. However, they often incur high execution costs by reasoning and exploring the physical world from scratch. To reduce these costs, we introduce RoboSkill, a framework that connects skill acquisition and reuse through an Explore, Execute, Evolve loop. Within this loop, the agent explores to gather task-relevant information, executes tasks while adapting to feedback, and evolves its skill library based on execution records. It then reuses these skills to guide exploration and execution in the next cycle, closing the loop. To improve loop efficiency, we complement vision with tactile feedback to reduce uncertainty during physical interaction. We further augment textual guidance with reusable code to reduce reasoning overhead during skill reuse. On LIBERO-10, RoboSkill improves first-episode success rates by 12.5--25.0 percentage points and reduces average runtime by 7.6--72.4% across four agents. On real robots, it improves success rates by 8.3 percentage points and reduces average runtime for successful trials by at least 14.4%.
Sicheng Xie, Yitong Chen, Haidong Cao +3
Institute of Trustworthy Embodied AI, Fudan University · Shanghai Innovation Institute · NeoteAI.