manfredch/lyapunov/robot-provisioning

Claude Code and agent skill by manfredch/lyapunov

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use for an explicit request to acquire/import a robot body or prepare a policy bundle. Reuse the body, complete cache, and Jobs; check required files and instance compatibility. Use action-execution for actions with an existing robot.

Install

npx vskill@latest install manfredch/lyapunov/robot-provisioning

SKILL.md

Acquire and prepare a robot

Use this flow only for acquisition, import, or policy preparation. Use action-execution to lower, grasp, or drive an existing robot. Verify visible geometry, complete bytes, preparation, compatibility, and achieved motion separately.

  1. Check for the same body with asset_list / scene_inspect; a human may import it from the robot library. Register original URDF/MJCF files and their include/mesh dependency closure with scene_import, verify files with asset_verify, and reuse resourceId/version and articulation/controller through scene_mount; do not duplicate the body.
  2. When a policy is needed, first read policy_load_state, policy_files cache:true, and related job_list / job_output. Reuse complete files. Read back an active download with the same identity, and acquire only missing files. Visible geometry does not prove that weights, adapters, or observation mappings are complete.
  3. External directories provide source metadata. Start from the body's official website or official project repository below; public source research does not require a regional service, proxy endpoint, or an existing runtime adaptation. When source, interface, or error facts are uncertain, use Browser Use for keyword search and official pages, WebFetch for a known URL, and Bash for necessary acquisition. Resolve new evidence before retrying the same failure. policy_download_sources lists registered pins/adapters/files only. For explicit authorized acquisition, report the actual source URL, retrieved revision, and missing files or requirements; retain authenticated-source permissions and complete-file verification. Manual policy_download_bundle/policy_download remain available with exact paths or directory prefixes ending in /, never wildcards or a copied historical matrix. Match Go1/Go2/G1, G1 12DOF, and 23DOF/75-dimensional joint/control/observation contracts from actual facts. Use only https://hf-mirror.com for HF; missing objects remain blocked without official-endpoint fallback.
  4. After complete acquisition, run policy_verify then policy_prepare; preparation returns derived artifacts and components/worldOptions. To apply them to the current instance, run policy_activate then policy_match, and enter action execution only after MATCHED. A download, PREPARED, or MATCHED does not prove standing or walking. Report exact missing runtime, original files, fixed source, or mapping and an executable way to resolve them.

Official source metadata, read on demand

These addresses cover the 12 documented T0 bodies in packs/t0-roster.json. T0 is body priority only. Select the exact model/version from the official source when needed; these roots do not assert a downloadable file, policy adapter, or supported behavior.

Body IDsOfficial source
allegro_handWonik Allegro Hand
crazyflie_2Bitcraze Crazyflie firmware project
forklift_cNVIDIA wheeled robot asset documentation
franka_pandaFranka Robotics ROS descriptions
generic_quadrotorgym-pybullet-drones upstream project
leap_handLEAP Hand project
shadow_handShadow Robot descriptions
unitree_a1, unitree_g1, unitree_go1, unitree_go2Unitree Robotics descriptions
ur10eUniversal Robots descriptions

ur5e uses the same official Universal Robots repository for source research; its T0/T1 documentation conflict remains unresolved. generic_quadrotor names a generic upstream model, and forklift_c names an NVIDIA asset family; inspect the actual selected body instead of inventing a manufacturer or matching variant.

Server directories supply official addresses and necessary context; they do not report actual robot behavior. Evaluate standing, gait, and getting up from the client's current simulation readback. A source match, download, PREPARED, MATCHED, or short inference does not verify these behaviors. Do not automatically inject remote capability/provisioning context for a target whose body/runtime/behavior adaptation has not been established. Peiri supplies the model; resolve public sources through the available Browser Use, WebFetch, and Bash tools.

Manual joint, gripper, and TCP control does not require a policy. Read real body/site/TCP/base data with robot_describe. Configure TCP with robot_set_tcp, and fixation/release/anchors with robot_set_base, including Scene revision/world generation. The human entry is the robot manual controls panel; read real presets with robot_presets when visible. A standard Panda has no site: explicitly choose a body origin/local offset rather than guessing a default TCP or using parentId as a weld.

Use scene_physics_update for properties and scene_bind_physics for derived bindings; use scene_reconcile_physics for pending or incorrect references. Handle robot bodies through their dedicated interfaces without overwriting internal joints. Read compiled physics after synchronization. Dependencies follow current tool configuration and isolated executor receipts; do not assume a system Python path, SDK version, or engine prerequisite.

Scene/Resource/Sim changes use domain interfaces, and workspace scripts follow existing permissions. Do not edit product source, state databases, or resource directories directly to bypass an interface. Save a portable project with scene_save portable. Deliver the acquired/missing/prepared/matched state for this task; verify motion separately.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/manfredch/lyapunov/robot-provisioning/badge)](https://verified-skill.com/skills/manfredch/lyapunov/robot-provisioning)

Something wrong with this skill or its page? Tell us.