estate-infrastructure

name	estate-infrastructure
description	Use when infrastructure work needs the current Estate map, resource state, declarations, provider readback, an exact plan, or the provider-native management loop. Trigger: /estate-infrastructure, /estate.
argument-hint	[observe-plan|manage]

/estate-infrastructure

Estate is the map and evidence surface for infrastructure context. This primitive teaches an agent to inspect that context and, when explicitly asked, operate the provider through the ordinary OpenTofu/provider tooling. Estate does not apply changes, mint credentials, or execute provider commands.

Navigate the live Estate surface

Read the live Estate revision selected by the workspace or control plane before making a recommendation or taking action:

1. Run estate map to find declared resources and their current Estate ids.
2. Run estate resource <id> for the selected resource's desired, observed, cost, health, recovery, and evidence projections.
3. Resolve the current repository by its exact owner_repo product identity, then follow Estate's graph relationships to its deployment, host, provider resources, health, cost, and recovery evidence. That graph is the routing table; do not copy Misty Step's private topology into a public product repository. A missing or ambiguous repository join is an evidence gap and keeps the agent read-only.
4. Read the linked declarations and the exact plan context for the requested scope. Do not infer resource identity, provider, or consequence from a role, pack, or remembered preference.
5. Read the provider readback and its observation time. Treat missing, stale, or contradictory projections as a stop condition, not an invitation to guess.
6. Read the current Estate standards, vendor defaults, applicable exceptions, and decisions that govern the resource. A standard with a passed review_date or an expired exception fails closed: report it and remain read-only until the evidence is refreshed.

Intent classes

• observe-plan: use Estate map/resource/declaration/provider-readback data for inventory, reconciliation, drift, cost, health, recovery, or forecasting. This intent may produce an exact OpenTofu/provider plan for review, but it never applies a plan, changes a provider, or requests a mutation credential.
• manage: use only after a saved exact OpenTofu/provider plan exists for the requested scope. Confirm the plan digest and resource bounds, obtain explicit operator approval of that exact plan and digest, record the named reviewer, obtain a scoped Mint/provider credential through the normal operator path, and apply with the existing OpenTofu/provider tooling. After apply, verify health and provider readback, then return a secret-free evidence pointer for Estate's next reconciliation. Do not fabricate an Estate write command or API.

High-risk or irreversible changes may require additional provider or organizational controls; they never weaken the base requirement for explicit operator approval of the exact plan. A role, pack, chat message, Powder card state, Roster authority-provider receipt, or Estate map entry cannot stand in for that approval; the approval names a human reviewer and the exact plan digest. Do not add an Estate execution or permission layer or an alternate mutation path; use the provider's ordinary tooling and its scoped credential instead.

Evidence and bundle hygiene

Keep literal credentials, provider snapshots, state, full plans, and raw logs out of Git and resolved bundles. Keep only safe references: the Estate revision and resource id, standards and exceptions read, exact-plan digest or external plan location, named operator review, provider-readback clock, health/cost/recovery summary, and the evidence pointer returned for Estate reconciliation. Never claim an apply, operator review, rollback, recovery, or healthy readback without live evidence.