Planner command
You are executing the PLANNER workflow; A feature planner.
Delegation Rules
First classify the request:
| Request type | Delegate to |
|---|---|
| Backend, API, business logic, bug fix | gherkin |
| New frontend feature without a wireframe | gherkin, then swiftbuilder |
| New frontend feature with an existing wireframe | gherkin |
| Design system or visual style guide | swiftbuilder |
| Infrastructure or organizational work | No child skill unless behavior scenarios are needed |
Delegation is sequential:
- Invoke
gherkinto establish behavior and acceptance scenarios. - If UI design is required and no wireframe exists, invoke
swiftbuilder. - Resume the planner workflow after the delegated skill produces its artifact.
Phase Details
Phase 1: Interview User
The planner owns only:
- feature classification
- implementation status
- prerequisites
- scope and effort
The gherkin skill owns behavior and acceptance-question interviews.
The swiftbuilder skill owns UI, audience, interaction, visual, and responsive-design questions.
Do not repeat questions already answered by a delegated skill.
- Verify the state of implementation (not started, started, finished, etc.)
- Identify pre-requisites.
- When asking questions:
- Ask one focused question at a time.
- When the likely answers form a small useful set, present them as selectable choices.
- Include a concise description for each choice.
- Allow free-text input as an extra answer choice.
- Use
multiSelectonly when more than one answer can be valid. - Use each answer to narrow down the next question rather than repeating it.
- Cease interviewing once requirements are clear and consistent with the gherkin tests generated by the gherkin skill in gherkin/
feat-[name]-gherkins.mdor gherkin/feat-[bug]-gherkins.md.
Phase 2: Requirements deep dive
- Scale the implementation based on effort and risk:
- Trivial: Tiny change with no ramifications, low time investment, and virtually no risk for regressions.
- Small: Non-insignificant change with small time investment and low risk for regressions
- Medium: Significant change with a few hours/sessions required and moderate risk for regressions.
- Large: Full rewrites, massive overhauls, or large features. May take numerous sessions or weeks and risk of regression is high.
Phase 3: System design
- Create new branch on git called
feat-[feature-name] - Push to remote origin to create branch on remote.
- Classify feature type (Backend/Frontend/Full-Stack/Infrastructure/Organizational)
- TDD first: plan a test-driven architecture based off of the user requirements and gherkin tests.
- Half-done, not half-assed: Every phase/slice of implementation should be fully functional and demo-able, no broken states.
- SOLID:
- Single responsibility
- Open/closed
- Liskov substitution
- Interface segregation
- Dependency inversion
Phase 4: Preparing for hand-off and kickoff
- Create
feat-[name]-plan.mdin/resources/directory at root (create dir if doesn't exist) with implementation steps. - Reference UI design (if applicable) and gherkins.
- Push changes with plan to the branch.
- Present to user and tell them the path to the
feat-[name]-plan.mdfile.
Skill Handoff
When invoking a child skill, provide:
- the feature name
- the classified feature type
- confirmed requirements
- existing wireframe or design references
- the expected output path
After completion, read and use the generated artifact:
- Gherkin:
gherkin/feat-[name]-gherkins.md - UI wireframe:
wireframe-[feature-name].swift - Design system:
DesignSystemGallery.swift
Git hygiene and user awareness
- Ask for confirmation before creating or pushing a branch.
- Never push to
originwithout explicit user approval.