xbot-dev

/xbot-dev — Start development on ready skills

Kick off the dev pipeline for all skills whose dependencies are satisfied.

Usage

/xbot-dev

What to do

1. Check prerequisites. All three services must be running:

curl -sf http://localhost:8766/entries  # orchestrator
curl -sf http://localhost:8080/state    # agent server

If the sim is needed (RoboCasa tasks), also check:

nc -z localhost 5500                    # sim

Also check the service catalog and remote server health:

curl -sf http://localhost:8090/health   # service catalog

If the service catalog is running, check:

• server_reachable — is the remote GPU server accessible?
• alert — any active alerts (server unreachable)?
• List any services with status: "server_down" or status: "stopped" that skills may depend on

If the catalog is not running, warn the user and suggest:

Run /xbot-server to set up the service catalog first.

If the remote server is unreachable, warn the user that GPU services (GraspGen, GroundedSAM, etc.) will not be available and skills depending on them will fail.

If any local service is down, tell the user what to start and stop.

1. Show the current tree. Fetch GET /entries and display it with statuses:

curl -s http://localhost:8766/entries | python3 -c "
import json, sys
entries = json.load(sys.stdin)
for e in entries:
    deps = ', '.join(e.get('dependencies', [])) or '(leaf)'
    status = e.get('status', 'planned')
    print(f'  [{status:>10}] {e[\"name\"]:30s} deps: {deps}')
"

1. Generate SKILL.md for each skill. Before starting dev agents, generate a SKILL.md for every skill in the graph that doesn't have one yet. For each skill:

   • Read the graph entry (name, description, dependencies)
   • Read the graph metadata (graph.json) for task context
   • If it has dependencies, read those dependency skills' SKILL.md files to understand what inputs/outputs they provide
   • Write SKILL.md to skills/<skill-name>/SKILL.md with this structure: