170 lines
7.1 KiB
Markdown
170 lines
7.1 KiB
Markdown
---
|
|
name: plan-ticket
|
|
description: Create a complete implementation plan for a HiveGrid ticket from AGENTS.md,
|
|
its ticket doc, linked planning sources, live Plane context, and current git
|
|
state. Use when a user provides a HiveGrid ticket ID or asks to analyze,
|
|
scope, sequence, or plan a ticket without implementing it yet.
|
|
---
|
|
|
|
# Plan Ticket
|
|
|
|
Plan only. Do not edit code, ticket docs, Plane, or external systems. Stop after
|
|
the plan and ask for approval before implementation.
|
|
|
|
## 1. Resolve the ticket
|
|
|
|
Require a ticket ID such as `AK-P3`. Then:
|
|
|
|
1. Read `AGENTS.md` completely enough to apply the current rules relevant to the
|
|
ticket.
|
|
2. Read `.planning/ticket-docs/README.md` and
|
|
`.planning/ticket-docs/<TICKET-ID>.md` completely. Stop if the ticket doc is
|
|
missing.
|
|
3. Follow only the granular sources linked by the ticket doc. Give
|
|
`.planning/ENTITY-MODEL.md` priority where `AGENTS.md` requires it.
|
|
4. Read the current branch, recent commits, and diff against `origin/main` to
|
|
establish what is already implemented.
|
|
5. Pull the live Plane ticket, comments, `blocked-by`, and `blocks` using
|
|
`.planning/plane-sync/fetch_ticket.py` or the available Plane integration. If
|
|
Plane is unavailable, continue from Git and the ticket doc but state that
|
|
evidence gap.
|
|
6. Never infer completion from a status badge alone.
|
|
|
|
## 2. Audit dependencies with three sources
|
|
|
|
Check both dependency directions. For every blocker or consumer, compare:
|
|
|
|
1. Plane state and all relevant comments.
|
|
2. Git/code reality, including merged commits and implementation notes.
|
|
3. The blocker and dependent ticket docs.
|
|
|
|
Decide whether the exact contract needed by this ticket exists. When sources
|
|
conflict, apply the precedence rules in `AGENTS.md`; otherwise expose the
|
|
conflict instead of inventing a resolution. For cross-cutting contracts,
|
|
schemas, shared services, or `blocks` consumers, include the dependent
|
|
verification work in the plan.
|
|
|
|
## 3. Decide the delivery strategy
|
|
|
|
Classify every open item as:
|
|
|
|
- `selbst machbar`
|
|
- `extern/blockiert`
|
|
- `bewusst v2/out-of-scope`
|
|
|
|
Recommend `jetzt bauen`, `teilweise bauen`, or `zurückstellen`, with a reason.
|
|
Do not pull work forward from later tickets.
|
|
|
|
For UI-heavy work, read `.planning/INTERIM-UI-PLAYBOOK.md` and
|
|
`.planning/INTERIM-UI-DESIGN-STANDARDS.md`. Missing final design does not
|
|
automatically block the ticket: plan the AD-28 interim screen when the ticket
|
|
permits it, and track final visual work as `design-nachzug`. Use finished
|
|
Automation/Forms screens and existing Settings Subscription/Billing interim
|
|
screens as style references.
|
|
|
|
If the ticket specifically integrates new or changed Subframe pages, also use
|
|
`$subframe-einbinden` and incorporate its inventory and integration
|
|
requirements.
|
|
|
|
## 4. Produce the plan in this order
|
|
|
|
1. Plain-language introduction
|
|
2. Current state and DoD percentage
|
|
3. Dependencies and delivery strategy
|
|
4. DoD-to-slices mapping
|
|
5. Ordered build plan by slice
|
|
6. Verification gates before commit
|
|
7. Open questions and risks
|
|
8. Reminder for post-build ticket-doc + Plane updates
|
|
|
|
## 5. Map every DoD item
|
|
|
|
For every DoD item, state:
|
|
|
|
- **Status:** done, open, or blocked.
|
|
- **Fertigkeitskriterium:** what must be true for full completion.
|
|
- **Proof:** the code path, test, screen, external artifact, or Plane evidence
|
|
that proves it.
|
|
- **Zählt NICHT als fertig:** which partial implementation would be
|
|
insufficient.
|
|
- **Slice:** where the work and verification happen.
|
|
|
|
Cross-check each DoD item against `Was wir bauen`, `Benötigte Screens`,
|
|
technical building blocks, Plane comments, and design notes. Count only fully
|
|
satisfied points. Report `% = completed DoD points / total DoD points`,
|
|
separating blocked work from work the developer can still do.
|
|
|
|
## 6. Make each slice executable
|
|
|
|
For each slice, name concrete files and include all applicable constraints:
|
|
|
|
- Loader/action builds the view model; UI receives VMs and callbacks.
|
|
- No `@prisma/client` or `~/db` in `packages/ui-app/**`.
|
|
- No direct `packages/ui-subframe` import from routes.
|
|
- No `useLoaderData` inside `packages/ui-app/**`.
|
|
- Reuse audit, activity, notifications, search, permissions, tenant scope, and
|
|
other cross-cutting services from `AGENTS.md`.
|
|
- Use i18n keys and the existing clear-text error/state patterns.
|
|
- For UI, include interim/final visual state, smoke, hot-swap, and
|
|
visual-baseline work as required.
|
|
- For schema work, use a module schema, develop with `prisma db push`, and plan
|
|
the current pre-merge migration/parity gates from `docs/AGENT-CHECKLIST.md`
|
|
without guessing around AD-21 timing.
|
|
|
|
Call out central registries whenever a slice adds or changes their inputs:
|
|
|
|
- Tenant-scoped Prisma model: tenant model registry plus ordering/tests.
|
|
- Workspace package subpath export: explicit Vitest alias plus consumer suites.
|
|
- Feature key: entitlements feature, blueprint, plan, and key-list tests.
|
|
- Permission key: permission seed, defaults, scope display, and de/en strings.
|
|
- Sidebar/navigation ID: UI registry, sidebar tests, and UI inventory mapping.
|
|
- UI text: de/en translations and i18n gates.
|
|
- Prisma schema: migration parity and the RLS live migration-chain gate.
|
|
|
|
## 7. Define verification before commit and push
|
|
|
|
Derive exact commands from `docs/AGENT-CHECKLIST.md`, the affected workspaces,
|
|
and `.gitea/workflows/ci.yml`. Include:
|
|
|
|
- typecheck, lint, and targeted tests for every affected package
|
|
- existing consumer suites for shared files, not only new tests
|
|
- applicable `verify:*` registry, i18n, RLS, migration, bundle, or Docker gates
|
|
- seed idempotency and browser/staging proof when the ticket requires them
|
|
- the no-US-provider and architectural anti-pattern checks when relevant
|
|
|
|
Do not schedule external writes, pushes, Plane changes, or Done transitions
|
|
without explicit authorization.
|
|
|
|
## 8. Include the post-build audit reminder
|
|
|
|
Before any completion report or Done-state recommendation, perform a
|
|
**Ticket-Abschluss-Audit**:
|
|
|
|
1. Re-read the ticket doc and current Plane comments.
|
|
2. Compare code state against every DoD item, required screen, and technical
|
|
building block.
|
|
3. Separate remaining work into `selbst machbar`, `extern/blockiert`, and
|
|
`bewusst v2`.
|
|
4. Count percentage only from fully satisfied DoD points. Partial backend-only
|
|
or UI-only work does not count as a full DoD point.
|
|
5. If a design ticket such as `D-*` is referenced, state whether its delta is
|
|
integrated, replaced by a valid interim baseline, or still blocking.
|
|
6. Update the ticket doc with date, proof, DoD checkboxes, and percentage before
|
|
any Plane completion action.
|
|
7. Mirror the result to Plane only with the required identity check and explicit
|
|
authorization; set Done only at 100% DoD.
|
|
8. If an interim screen was created, offer the optional David handover described
|
|
in `.planning/INTERIM-HANDOVER-GUIDE.md` instead of creating it silently.
|
|
|
|
End the response after the plan and ask whether it should be approved or
|
|
adjusted.
|
|
|
|
## Mandatory constraints
|
|
|
|
- Follow `AGENTS.md` strictly.
|
|
- Stay in ticket scope.
|
|
- Keep generated Subframe app chrome out of app screen components.
|
|
- Mark out-of-scope ideas explicitly as
|
|
`💡 Vorschlag (außerhalb Ticket-Scope): ...`.
|
|
- If a source is ambiguous, stop and ask instead of inventing.
|