agentteams-worker

Repo	box/github.com/opencharly/pod-agentteams-worker:v2026.274.0808

The AgentTeams Worker agent image — the openclaw gateway runtime the controller spawns for every Worker CR. Composes the shared agentteams-openclaw candy (openclaw built from the pinned upstream source commit) and the shared agentteams-cli candy (agt REST client + mc), then installs the shared AgentTeams protocol libs (/opt/agentteams/scripts/lib/) from the pinned AgentTeams source. The charly-owned start-worker.sh sources those libs, pulls the worker config (openclaw.json, SOUL.md, AGENTS.md, skills) from MinIO, starts the file-sync loops, re-logins to Matrix for a fresh E2EE device, and execs the openclaw gateway. The controller passes HOME=/root/agentteams-fs/agents/ (hardcoded upstream) and the worker’s config/state all lives in MinIO (stateless), so the image chmods /root to 755 (traversable by uid 1000) and pre-creates /root/agentteams-fs + /root/.openclaw uid-1000-writable at build time; the per-worker agents/ workspace is created at runtime. Rootless: the service runs as the image user (uid 1000).

Services

• agentteams-worker

Acceptance plan

This candy’s plan: — the runnable spec charly check executes against a live deployment. check: steps are idempotent probes; run: steps change state.

Intent	Step
run	install the shared AgentTeams protocol libs
run	pre-create the rootless worker runtime directories
run	write the rootless worker start script
check	the shared AgentTeams protocol libs are installed
check	the worker start script is installed
check	the worker runtime directories are uid-1000-writable and /root is traversable