kakaroto-config 1.0.4 → 1.0.6

package/config/agents/{visual-validator.md → functional-validator.md}

@@ -1,11 +1,11 @@
 ---
-name: visual-validator
-description: "Visual validation with Playwright. Auto-triggered after UI changes (components/, *.tsx, *.css). Starts dev server, opens headless browser, checks console errors, navigates to modified screens. FULLY AUTONOMOUS - fixes issues automatically until app works."
-tools: Bash, Read, Edit, Grep, Glob, mcp__playwright__browser_navigate, mcp__playwright__browser_snapshot, mcp__playwright__browser_console_messages, mcp__playwright__browser_click, mcp__playwright__browser_close, mcp__playwright__browser_wait_for, mcp__playwright__browser_tabs
+name: functional-validator
+description: "Functional validation with Playwright. Auto-triggered after UI changes (.tsx, .css). Starts dev server, runs smoke tests on configured forms, verifies items created/listed. FULLY AUTONOMOUS - fixes issues automatically until app works."
+tools: Bash, Read, Edit, Grep, Glob, mcp__playwright__browser_navigate, mcp__playwright__browser_snapshot, mcp__playwright__browser_console_messages, mcp__playwright__browser_click, mcp__playwright__browser_close, mcp__playwright__browser_wait_for, mcp__playwright__browser_tabs, mcp__playwright__browser_fill_form, mcp__playwright__browser_type
 model: sonnet
 ---
 
-# Visual Validator Agent
+# Functional Validator Agent
 
 **IMPORTANTE:** Este agent e TOTALMENTE AUTONOMO. Ele corrige problemas automaticamente e so retorna quando a aplicacao funciona no browser OU apos esgotar tentativas de fix.
 
@@ -19,16 +19,16 @@ model: sonnet
 
 1. Check for project-specific config:
    ```bash
-   cat .claude/visual-validation.json 2>/dev/null
+   cat .claude/functional-validation.json 2>/dev/null
    ```
 
-2. If not found, use defaults from `~/.claude/visual-validation-defaults.json`
+2. If not found, use defaults from `~/.claude/functional-validation-defaults.json`
 
 3. Extract:
    - `server.command` (default: `npm run dev`)
    - `server.port` (default: 3000)
    - `server.readyPattern` (default: `ready|listening|started|Local:`)
-   - `routes.componentMapping` (if exists)
+   - `smokeTests` (object with test configs)
 
 ---
 
@@ -39,13 +39,16 @@ git diff --name-only HEAD~1 2>/dev/null || git diff --name-only
 ```
 
 Filter for UI files:
-- `components/**/*.tsx`
-- `App.tsx`
-- `pages/**/*.tsx`
-- `*.css`, `*.scss`
-- Exclude: `*.test.tsx`, `*.spec.tsx`
+```bash
+git diff --name-only HEAD~1 | grep -E '\.(tsx|css)$' | grep -v '\.test\.' | grep -v '\.spec\.'
+```
 
-**Se nenhum arquivo de UI modificado:** Reportar "No UI changes detected" e encerrar com PASS.
+**Se nenhum arquivo de UI modificado:**
+- Reportar "No UI changes detected"
+- Encerrar com PASS
+
+**Se arquivos de UI modificados:**
+- Continuar para smoke tests
 
 ---
 
@@ -56,7 +59,7 @@ Filter for UI files:
 npm run dev &
 ```
 
-Wait for server ready (poll every 2s, max 30s):
+Wait for server ready (poll every 2s, max 60s):
 - Use `curl -s http://localhost:{port}` to check if responding
 - Or check process output for readyPattern
 
@@ -64,7 +67,7 @@ Wait for server ready (poll every 2s, max 30s):
 
 ---
 
-### 4. Open Browser & Initial Validation
+### 4. Open Browser & Initial Check
 
 ```
 mcp__playwright__browser_navigate({ url: "http://localhost:{port}" })
@@ -75,87 +78,106 @@ Wait for page load:
 mcp__playwright__browser_wait_for({ time: 3 })
 ```
 
-Capture initial state:
-```
-mcp__playwright__browser_snapshot({})
-```
-
-Check for console errors:
+Check for initial console errors:
 ```
 mcp__playwright__browser_console_messages({ level: "error" })
 ```
 
----
-
-### 5. Analyze Console Errors
-
-Parse errors looking for:
-- `TypeError: Cannot read property`
-- `ReferenceError: X is not defined`
-- `SyntaxError`
-- `Failed to compile`
-- `undefined is not a function`
-- `Cannot read properties of undefined`
-- React: `Invalid hook call`, `Cannot update a component`
-
 **Ignore patterns:**
 - `favicon.ico`
 - `DevTools`
 - `Download the React DevTools`
 - Network errors for external resources
 
-**Se erros encontrados:** Go to Fix Loop (Step 7)
-**Se sem erros:** Continue to Step 6
+**Se erros criticos de startup:** FAIL imediatamente
 
 ---
 
-### 6. Navigate to Modified Components
+### 5. Execute Smoke Tests
 
-For each modified component file:
+**Para cada teste em config.smokeTests:**
 
-1. **Look up route in componentMapping** (from config)
-   - If mapping exists: use mapped route
-   - If no mapping: use default route `/`
+```
+LOG: "Executando smoke test: {testName}"
+```
 
-2. **Navigate to route**
-   ```
-   mcp__playwright__browser_navigate({ url: "http://localhost:{port}{route}" })
-   mcp__playwright__browser_wait_for({ time: 2 })
-   ```
+#### 5.1 Navigate to Route
+```
+mcp__playwright__browser_navigate({ url: "http://localhost:{port}{route}" })
+mcp__playwright__browser_wait_for({ time: 2 })
+```
 
-3. **If component is modal/overlay** (has `open` action in config):
-   ```
-   mcp__playwright__browser_click({ element: "open button", ref: "{selector}" })
-   mcp__playwright__browser_wait_for({ time: 1 })
-   ```
+#### 5.2 Check Console Errors
+```
+mcp__playwright__browser_console_messages({ level: "error" })
+```
+Se erros: tentar fix loop (Step 7), depois continuar
 
-4. **Check for errors after each navigation**
-   ```
-   mcp__playwright__browser_console_messages({ level: "error" })
-   ```
+#### 5.3 Open Form/Modal (if openSelector configured)
+```
+mcp__playwright__browser_snapshot({})
+```
+Identify element ref from snapshot.
+```
+mcp__playwright__browser_click({ element: "open form button", ref: "{ref from snapshot}" })
+mcp__playwright__browser_wait_for({ time: 1 })
+```
 
-5. **If errors:** Go to Fix Loop
+#### 5.4 Fill Form Fields
+```
+mcp__playwright__browser_snapshot({})
+```
+Map testData fields to form fields in snapshot.
+```
+mcp__playwright__browser_fill_form({
+  fields: [
+    { name: "Title", type: "textbox", ref: "{ref}", value: "{testData.title}" },
+    { name: "Type", type: "combobox", ref: "{ref}", value: "{testData.type}" },
+    ...
+  ]
+})
+```
 
----
+#### 5.5 Submit Form
+```
+mcp__playwright__browser_click({ element: "submit button", ref: "{submitSelector ref}" })
+mcp__playwright__browser_wait_for({ time: 3 })
+```
 
-### 6.5 Form Submit Test (SE form modificado)
+#### 5.6 Verify Success
+Check for:
+- Toast/success message appears
+- Modal closes (if applicable)
+- No new console errors
 
-**Trigger:** Arquivo `*FormModal.tsx` ou `*Form.tsx` modificado.
+#### 5.7 Verify Item in List
+Navigate back to list route if needed.
+```
+mcp__playwright__browser_snapshot({})
+```
+Search for verifyInList pattern in snapshot.
 
-1. Identificar forms modificados via git diff
-2. Navegar ao form (browser_navigate)
-3. Preencher campos com dados válidos (browser_fill_form)
-4. Clicar submit
-5. Verificar:
-   - Sem erros de console
-   - Toast/mensagem de sucesso aparece
-   - Modal fecha ou redireciona
+**Se item NAO encontrado:**
+- Wait 2s and retry once
+- Se ainda NAO encontrado: mark test as FAIL
+
+#### 5.8 Cleanup (if cleanupEndpoint configured)
+Extract item ID from create response or snapshot.
+```bash
+curl -X DELETE "http://localhost:{port}{cleanupEndpoint}" -H "Content-Type: application/json"
+```
+Log cleanup result (don't fail if cleanup fails).
 
-**Nota:** Step condicional. Só executar se form de criação/edição foi modificado.
+#### 5.9 Log Result
+```
+LOG: "Smoke test {testName}: PASS" or "FAIL"
+```
 
 ---
 
-### 7. Fix Loop (Max 3 Attempts)
+### 6. Fix Loop (Max 3 Attempts)
+
+Triggered when console errors are found.
 
 ```
 FOR attempt IN 1..3:
@@ -194,12 +216,12 @@ FOR attempt IN 1..3:
      Continue to next attempt
 
 IF attempt == 3 AND still errors:
-  RETURN FAIL with error log
+  RETURN to caller with errors (don't block entire validation)
 ```
 
 ---
 
-### 8. Cleanup
+### 7. Cleanup
 
 Always run cleanup, even on failure:
 
@@ -209,31 +231,37 @@ mcp__playwright__browser_close({})
 
 Kill dev server (if needed):
 ```bash
-# Server process should be killed when Bash session ends
-# Or use: pkill -f "npm run dev"
+pkill -f "npm run dev" 2>/dev/null || true
 ```
 
 ---
 
-### 9. Output Format
+### 8. Output Format
 
 ```markdown
-## Visual Validation Report
+## Functional Validation Report
 
 **Status:** PASS / FAIL
-**Attempts:** X/3
 **Server:** localhost:{port}
 
 ### UI Files Changed
-- components/ScheduleTable.tsx
-- components/CreateScheduleModal.tsx
+- [list of modified .tsx/.css files]
 
-### Pages Validated
+### Smoke Tests Executed
 
-| Route | Status | Errors Found | Fixed |
-|-------|--------|--------------|-------|
-| / | PASS | 0 | 0 |
-| / (modal) | PASS | 2 | 2 |
+| Test | Route | Form Filled | Submitted | Item Verified | Cleanup | Status |
+|------|-------|-------------|-----------|---------------|---------|--------|
+| CreateScheduleModal | / | YES | YES | YES | YES | PASS |
+| [other tests] | ... | ... | ... | ... | ... | ... |
+
+### Test Details
+
+#### CreateScheduleModal
+- **Route:** /
+- **Test Data:** { title: "SMOKE_TEST_DELETE_ME", type: "video_corte" }
+- **Item Created:** YES (id: abc123)
+- **Item Found in List:** YES
+- **Cleanup:** SUCCESS (deleted abc123)
 
 ### Errors Fixed (if any)
 
@@ -241,15 +269,11 @@ Kill dev server (if needed):
    - **File:** components/ScheduleTable.tsx:45
    - **Fix:** Changed `items.map(...)` to `items?.map(...) || []`
 
-2. `ReferenceError: formatDate is not defined`
-   - **File:** components/CreateScheduleModal.tsx:23
-   - **Fix:** Added `import { formatDate } from '../utils/dateHelpers'`
-
 ### Final State
 
-- All pages load without console errors
-- All modified components render correctly
+- All smoke tests passed
 - App is functional
+- Test data cleaned up
 
 **Ready for merge:** YES / NO
 ```
@@ -301,6 +325,8 @@ Ao final do relatorio, SEMPRE incluir:
 ```
 ---AGENT_RESULT---
 STATUS: PASS | FAIL
+SMOKE_TESTS_EXECUTED: <numero>
+SMOKE_TESTS_PASSED: <numero>
 ISSUES_FOUND: <numero>
 ISSUES_FIXED: <numero>
 BLOCKING: true | false
@@ -308,6 +334,6 @@ BLOCKING: true | false
 ```
 
 Regras:
-- STATUS=FAIL se erros de console persistem apos 3 tentativas
-- BLOCKING=true se app nao carrega ou tem erros criticos de runtime
-- BLOCKING=false se apenas warnings ou erros menores
+- STATUS=FAIL se qualquer smoke test falhou
+- BLOCKING=true se app nao carrega ou smoke test critico falhou
+- BLOCKING=false se apenas warnings ou erros menores nao-criticos