tidybot-run-robot-task

name	tidybot-run-robot-task
description	Execute a physical task on the robot (pick up, place, move, look at, etc.). Use when the user asks you to do something with/on the robot — any manipulation, navigation, or perception task. Triggers on requests like "pick up the banana," "put that on the plate," "go to the table," "what do you see," etc.

Run Robot Task

When the user asks you to do something on the robot, follow this order strictly. Do not skip to writing raw SDK code.

Step 1: Check Existing Skills

1. Check dev/ folder for local in-progress skills that match the task
2. Fetch the skill catalog: GET http://<ROBOT_IP>:8080/skills/catalog (or check your memory for known skills)
3. If a matching skill exists, use it — do not reinvent

Step 2: Chain If Possible

If no single skill matches but multiple existing skills can be combined:

1. Break the task into sub-steps mapped to existing skills
2. Execute them in sequence, passing context between steps
3. Chaining tested skills beats untested new code

Step 3: Build New Only as Last Resort

If no existing skill covers the task:

1. Read the SDK guide first: GET http://<ROBOT_IP>:8080/docs/guide/html
2. Check the active-services skill for available backends
3. Follow the tidybot-skill-dev skill to build, test, and save to dev/ (use tb- prefix for robot skills)

Execution Checklist

Before running any code on the robot:

• Acquire a lease (POST /lease/acquire)
• Read the SDK guide if writing new code (every session, no exceptions)
• Use print() for status — poll /code/status for output
• Check recorded frames after execution to verify results
• Release lease or let it expire when done

Debugging Failures

When execution fails or hangs:

1. Check /code/status — includes stderr with tracebacks
2. Check ~/tidybot_uni/logs/ — persistent logs for everything:
   • agent_server.log — server runtime (connections, bridge errors, lifecycle)
   • code_execution.log — stdout/stderr from all executions
   • code_executions/ — per-execution saved scripts + output
3. Do NOT redirect server output to /tmp/ — the built-in logs already capture everything

Key References

• Robot connection: see robot-connection skill
• Hardware specs: see robot-hardware skill
• SDK methods: see robot-sdk-ref skill
• Building new skills: see tidybot-skill-dev skill
• Available services: see active-services skill