using-loomlisted

# using-loom <EXTREMELY-IMPORTANT> If you think there is even a 1% chance a skill might apply, you ABSOLUTELY MUST invoke the skill. IF A SKILL APPLIES TO YOUR TASK, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT. This prevents silent scope invention, retroactive tickets, unbounded workers, unsupported closure claims, and lost recovery context. </EXTREMELY-IMPORTANT> Loom is a human-agent control plane. Its durable state is the `.loom/` graph: Markdown records for judgment, bounded execution, evidence, audit, and reusable knowledge. Chat shapes the next move; records preserve what future agents need. ## Loop Order Loom routing comes first. Activation is part of routing: if any Loom skill or surface might own the next move, invoke the relevant skill before responding, before asking clarifying questions, before code exploration, before quick checks, before editing files, before creating tickets, and before launching Ralph. Use workflow playbooks only after routing identifies the owning surface. The default sequence is: ```text shape with the operator -> route durable truth -> slice executable work -> execute ticket slices through bounded Ralph runs -> preserve evidence -> audit claims -> reconcile records ``` Ambiguity defaults to shaping, not implementation. Execute only when outcome, scope, constraints, success criteria, evidence posture, non-goals, and material system-shape or state implications are clear enough to avoid hidden product or architecture choices. Otherwise