audience: AI coding agents first.

# Factory supervision and resolution — execution manifest

**Goal:** route each advisor-approved implementation request through its correct trust domain without letting factory edit its judge.

**Canonical design:** `docs/specs/2026-08-09-factory-supervision-and-resolution-design.md`.

## Requests

1. `docs/plans/2026-08-09-factory-supervisor-producer.md` — trusted independent agent; NEVER factory.
2. `docs/plans/2026-08-09-incident-resolution-coordinator.md` — trusted independent agent; NEVER factory.
3. `docs/plans/2026-08-09-incident-resolver-defaults-api.md` — factory after incident selector + trusted coordinator prerequisites.
4. `docs/plans/2026-08-09-factory-resolution-ui.md` — factory after Astryx, `DiffView`, APIs, and trace seams.
5. `docs/plans/2026-08-09-factory-resolution-integration.md` — trusted independent agent after all tracks; NEVER factory.

## Contract

Shared types and invariants come only from canonical design. Dependent requests copy exact seam names; they MUST NOT redefine them.

## Behavior

Execute trusted tracks outside factory. Launch factory-owned requests only after their listed prerequisites land. Install local infrastructure live before landing. Run integration canary last.

## Out of scope

This manifest contains no implementation body or task graph. NEVER pass this aggregate manifest to one factory run.

## Acceptance

All five request receipts are green; provider-free integration canary proves supervised success, independent repair, incident resolution, and owner contract decisions through live installed entrypoints and UI.