stetsom-front

Read the codebase knowledge graph for Stetsom Front. The graph lives in graphify-out/ and was generated by Claude Code. Use it for codebase orientation, impact analysis, and dependency tracing. VSCode agents cannot regenerate the graph — use /graphify in Claude Code CLI for that.

Full-site Figma fidelity audit for Stetsom Front. Downloads all page frames (desktop + mobile), performs cross-page divergence analysis with a ranked severity matrix, implements phased corrections, validates with tsc+lint, and exports baseline images. Use for periodic design audits, before major releases, or after big refactors. Trigger keywords: design audit, full design review, fidelity check, design fidelity, figma audit, audit figma, all pages review, full site audit.

Full-site Figma fidelity audit for Stetsom Front. Downloads all page frames (desktop + mobile), performs cross-page divergence analysis with a ranked severity matrix, implements phased corrections, validates with tsc+lint, and exports baseline images. Use for periodic design audits, before major releases, or after big refactors. Trigger keywords: design audit, full design review, fidelity check, design fidelity, figma audit, audit figma, all pages review, full site audit.

Audits screens/flows against Figma for completeness, correctness and consistency. Accepts any scope: a specific screen name, area ("cms", "admin", "website", "produtos"), or nothing (audits everything). Produces a ranked divergence matrix of missing flows, wrong fields, incorrect columns, broken UX, data model mismatches. Always presents the matrix to the user BEFORE implementing anything. Trigger keywords: screen audit, tela audit, audit screens, audit cms, audit website, audit flows, fluxo faltando, tela errada, validar telas, revisar telas, checar telas, missing screens, screen review, ux audit.

Generate a copy-paste ready git commit message from staged changes (or unstaged changes if nothing is staged). Read-only — never stages, never commits. Triggers: commit, make commit, quick commit, git commit, create commit, commit message, generar commit.

Refactor large or redundant code into modular, reusable components following Stetsom stack conventions (Next.js App Router, React 19, Tailwind v4, shadcn/ui). Use when files are too large, logic is duplicated, or components lack proper separation of concerns. Triggers: modularize, componentize, refactor, split component, extract component, reduce redundancy, cleanup code.

Design refinement pass using Figma MCP. Fetches live design from Figma (never local files), maps tokens, validates code against design standards, updates local context if stale, and produces an action plan via /next-task. Trigger keywords: refine design, figma alignment, design pass, design review, check design.

Use when the implementation of a feature or bugfix is complete and you need to generate a professional, conventional pull request description summarizing the changes against the main branch. Also updates the related task status to DONE in docs/ia/tasks/ and appends a changelog entry to docs/ia/context.json.

Use after brainstorm is complete — turns the refined idea into a tracked task with a git branch, acceptance criteria, and implementation plan. Trigger keywords: create task, add task, track task, new task.

Full-cycle delivery orchestrator — runs next-task, create-pr, code-review, fixes issues and repeats until review passes. Trigger keywords: deliver, full cycle, implement and ship, next delivery.

Use to start or resume implementation of the next available task — picks the next TODO task, reads its acceptance criteria, and drives implementation to completion. Trigger keywords: next task, start task, implement task, execute task.

Use when starting a new feature or when the direction is unclear — refines the idea, clarifies the objective, defines scope, and produces acceptance criteria ready for task creation. Trigger keywords: brainstorm, refine idea, clarify feature, define scope, plan feature.

Use when a feature branch is complete and ready for PR — runs a full diff-based review of the current branch against main, reads every changed file, validates against project rules, and produces a concrete action plan. Do NOT use on main branch.

Create or update agent rules for all runtimes using the dual-install pattern: .agents/rules/<topic>.md (VS Code Copilot Chat) + .claude/rules/<topic>.md (Claude Code CLI). Both folders are auto-discovered — no edits to AGENTS.md or .github/copilot-instructions.md needed. Use when adding coding conventions, framework rules, file-type standards, or project-specific guidelines every agent must follow. Triggers: "add rule", "create rule", "add project rule", "add coding conventions", "add project guidelines", "agent rules", "project rules".

Create a new project-specific agent skill and install it in both .agents/skills/ (VS Code Copilot) and .claude/skills/ (Claude Code CLI), following the skills.sh dual-install pattern. Use when adding custom rules, domain knowledge, or reusable workflows for all agents in the project. Triggers: "create skill", "add skill", "new skill for all agents", "project-specific rules", "add agent rules".