# Implementation plan: [app name / next change]

## Starting point
Product brief and architecture: [links + current versions]
Repository/baseline: [exact version; unknown if not inspected]
Run instructions and safe test data: [how to verify the baseline]
Verified behavior: [what currently works and how it was checked]
Preserve: [journeys, records, interfaces, and product rules]

## First complete outcome
[One user starts from X, completes Y, and can verify Z.]
Scope includes its UI, saved state, access rules, and important failure path.

## Milestone [number]: [observable outcome]
Depends on: [necessary preceding outcomes]
Affected boundaries: [architecture responsibilities]
Code locations: [inspect and identify; do not invent paths]
Work: [connected implementation steps]
Acceptance:
- Given [state], when [action], then [observable result].
- Reopen and verify [saved/delivered state].
- As [other role], verify [allowed/denied action].
- On [conflict/retry/failure], verify [safe user-visible behavior].
Evidence: [version, environment, commands/journey, observed result]
Done when: [all acceptance checks pass; limitations are recorded]

## Remaining milestones and dependencies
[Outcome | depends on | preservation duties | acceptance evidence]
Use complete outcomes rather than a list of screens or files.

## Decision and change boundaries
Implementer may decide: [routine technical/design choices]
Builder must resolve: [consequential product rules and commitments]
If inspection changes the plan: [finding, affected outcome, revised work]
Preserve confirmed intent; update the plan when implementation reality changes.

## Release gate
[Real-environment persistence, access, conflicts, service failures,
migrations, deployment, recovery, and remaining operating duties]
Owner: [who verifies and releases]

## Progress and evidence
[Milestone | not started/in progress/verified | evidence | remaining limits]
Files changed or model confidence alone do not prove completion.