Team Runtime AGENTS Template (Single Template)

Use this single template for all Team members. Keep runtime AGENTS.md minimal and role-scoped to control context size.

Agent Profile

member_id:
role: coordinator|worker
current phase: team formation|task analysis|role assignment|communication and collaboration|consensus formation|result integration
transition condition:

shared baseline (always):
- team-agents-index
- team-actor-mailbox
role core (pick one):
- coordinator: team-coordinator-orchestrator
- worker: team-worker-executor
optional (load only when needed):
- team-task-lifecycle
- team-deliberation-rules

Shared routing, mention, and human-facing reply rules live in skills/team/AGENTS.md.
Coordinator-specific execution policy lives in team-coordinator-orchestrator.
Worker-specific execution policy lives in team-worker-executor.
Mailbox transport and payload details live in team-actor-mailbox.
Canonical Team task creation/state rules live in team-task-lifecycle.
Self-profile updates use profile_patch_proposal.
Timed self-reminders use agenthub actor time-trigger-set, agenthub actor time-trigger-list, and agenthub actor time-trigger-cancel.
Operator-controlled idle reminders use agent_loop; treat them as follow-up nudges, not new human intent.
Load team-deliberation-rules only when you need option comparison or consensus work.

TODO.md
.agenthubmemory/TODO.md (worker/project workspace memory ledger, when applicable)
.agenthubmemory/journal/ (worker/project work log)
.agenthubmemory/note/ (worker/project durable lessons / heuristics)
.cache/context/ runtime continuity artifacts (not a durable TODO source)
latest evidence/log paths:

status: pending|in_progress|completed|blocked
latest update:
latest finding:
primary route: coordinator-mailbox|peer-mailbox|shared-channel (direct mailbox first unless the update needs shared visibility)
secondary route: none|human-notification
next checkpoint:
overdue action: