Skill: Canonical orientation

Use this as the boot skill for CairnStone work.

1. Read the canonical docs/AI_OPERATING_GUIDE.md referenced by the provider/project instructions.
2. Call cairnstone_health and note the live version and tool catalog; do not assume a cached connector schema is current.
3. Resume cairnstone-v6-project-memory with cairnstone_resume_chain(detail="start_here") for normal continuation. Treat the returned chain HEAD and sparse-authority identity as accepted orientation; never infer currentness from timestamps. Expand to detail="compact" only when accepted path-head/HEAD-edge context is needed, and to detail="full" only for explicit complete authority/edge inspection. If the server advertises detail=start_here but the client schema does not expose detail, report the schema mismatch and use the safest available fallback until the connector is refreshed.
4. Check both canonical AC1 inbox planes for the current main-session model when the client exposes correspondence tools: <namespace>:chat for conversation/coordination/design and <namespace>:cairnstone-v6 for durable repo/engineering handoffs and existing work history. Inspect compact inbox metadata first and read only relevant unread messages; inbox LOD5 is coordination metadata, not accepted project state. Do not replace the durable work inbox with a :cairnstone-v7 alias merely because the runtime is V7. Compatibility aliases are supplemental only. Correspondence transports coordination intent but never grants execution authority.
5. For multi-chain retrieval, use cairnstone_vault_catalog → cairnstone_resolve_scope → cairnstone_find_scope. Scope is retrieval/navigation context, never synthetic global authority. Do not assume cairnstone_ask_scope exists until live discovery and accepted documentation both confirm it is shipped.
6. Load specialized skills only after orientation. Prefer cairnstone_resolve_skills and then cairnstone_get_skill for the smallest useful set.

Do not mutate project state during orientation.