/frontend

/pwdev-feat:frontend — Create Frontend Plan

Agent

Assume the persona of agents/agent-planner.md (PWDEVIA).

Model Resolution

Read .planning/config.json for model_profile and model_overrides. Resolution order: (1) model_overrides[agent-name] → (2) profile lookup → (3) agent frontmatter model: default. Profiles — performance: opus for all except reviewer/scanner (sonnet). balanced: opus for orchestrator, sonnet for planner/executor/builder/interviewer/reviewer/researcher, haiku for scanner. economy: sonnet for most, haiku for reviewer/scanner. When spawning the agent, pass the resolved model via the model parameter.

Input

$ARGUMENTS: frontend task description (required).

Plan Type

Frontend — components, pages, composables/hooks, styles, E2E tests.

Flow

STEP 0 — Language Selection

Read .planning/config.json for the lang field (pt-BR or en). If set → use it silently. If not set → detect from $ARGUMENTS or ask: "Em qual idioma deseja seguir? / Which language would you like to use? 1. Portugues (PT-BR) 2. English (EN)" Save choice to .planning/config.json (merge, do not overwrite other fields). All subsequent output follows the resolved language. Technical terms stay in English.

1. Read context: CLAUDE.md + .planning/feat/codebase.md
2. Interpret the frontend request
3. Ask clarifying questions if needed (max 2 rounds)
4. Apply the 7 fundamental questions with frontend focus:
   • Persona: frontend engineer (Vue, React, Svelte, etc.)
   • Quality: component architecture, TypeScript, accessibility
   • Must include: Playwright E2E tests for all UI features
   • Must include: all states (loading, empty, error, success)
5. Create feature directory and generate plan: mkdir -p .planning/feat/features/{slug} then write .planning/feat/features/{slug}/plan.md
6. Present summary with next step: /pwdev-feat:exec {slug}

Prohibitions

• NEVER write code — only the plan
• NEVER skip Playwright E2E tests when there is UI
• NEVER plan components without defining all states