Working inside a Stoat coordination plan Added in 1.7.0
Work as an agent inside a Stoat coordination plan: find ready work with next_ready_nodes, claim/report/release nodes over MCP, correlate test runs with BOBCAT_RUN_TAG, prove a test-run gate and each of its declared steps with stoat gate exec (exactly the declared command), record a review node's findings with submit_review, never merge your own PR whatever the node's merge policy, ask a question with ask_person (marked for the plan's lead agent or for a person) instead of guessing, lead a plan run as the first responder for its workers' questions (lead_plan / await_questions / answer_question / escalate_question), park on await_dependencies / await_package_version / await_memories instead of polling, recall what earlier sessions learned (recall, get_repository) before starting, and record what the next session should know (record_memory, add_repository_note).
Use when asked to pick up work from a coordination plan, coordinate with other agents across repositories, wait on an upstream publish, or remember something across sessions.
Skill details
- Skill name:
stoat-coordination-agent-workflow - Applies to:
stoat - Tags:
stoat,coordination,agents,mcp,planning,github,nuget,ci,memory,recall,knowledge-base,questions,lead-agent - Added in: 1.7.0
- License: Proprietary. Copyright JasperFx Software LLC.