@liquidmetal-ai/raindrop 0.6.4 → 0.8.0

package/templates/goose/debug-raindrop-app.yaml

@@ -0,0 +1,91 @@
+version: "1.0.0"
+title: "Debug Existing Raindrop Application"
+description: "Debug an existing deployed Raindrop application"
+
+instructions: |
+  WARNING: If you cannot find the "raindrop-mcp" server or the "get-prompt" tool for the "raindrop-mcp" server, respond with:
+  "Please check your raindrop-mcp server status by running /mcp, and reauthenticate"
+
+  Before debugging, run `raindrop auth login` to authenticate. When the command returns, authentication is complete.
+
+  After authentication:
+
+  1. If session_id is provided as an argument, use it and skip to step 5
+
+  2. Otherwise, read `~/.raindrop/index.json` to get applications and sessions:
+     - File format:
+       ```json
+       {
+         "applications": {
+           "my-app": {
+             "latest_session_id": "sess_abc123",
+             "sessions": [...]
+           }
+         },
+         "sessions": {
+           "sess_abc123": {
+             "session_id": "sess_abc123",
+             "timeline_id": "timeline_456",
+             "created_at": "...",
+             "current_state": "...",
+             "application_name": "my-app"
+           }
+         }
+       }
+       ```
+
+  3. Handle application selection:
+     - If no applications exist, respond: "No Raindrop applications found. Use /new-raindrop-app to create one."
+     - If one application exists, use its `latest_session_id`
+     - If multiple applications exist:
+       * List all application names with their current state and last updated time
+       * Ask user: "Which application would you like to debug?"
+       * Use the selected application's `latest_session_id`
+
+  4. If no session_id found, ask the user for it
+
+  5. Call the `get-prompt` tool (from raindrop-mcp server) with session_id to rehydrate session context
+
+  6. Spawn a diagnostic subagent with this prompt:
+     "You are debugging a deployed Raindrop application.
+
+     Session ID: [session_id from step 5]
+     Timeline ID: [timeline_id from step 5]
+     Application Name: [application_name from step 5]
+     Application Type: [application_type from step 5]
+
+     Your task:
+     1. Ask user: 'What issues are you experiencing with your deployed application?'
+     2. Gather diagnostic information:
+        - Run: raindrop logs [application_name] --tail 100
+        - Based on application_type, test endpoints/tools:
+          * For API: Test endpoints from interface_design
+          * For MCP: Test MCP tools
+          * For CLI: Test commands
+        - Record failures and error messages
+     3. Aggregate test results into this format:
+        {
+          total_endpoints: X,
+          passing: Y,
+          failing: Z,
+          failure_details: ['endpoint/tool: error message', ...]
+        }
+     4. Call the `update-state` tool (from raindrop-mcp server) with:
+        - session_id: [session_id]
+        - timeline_id: [timeline_id]
+        - artifacts: {
+            current_state: 'debug_analyze',
+            is_debugging: true,
+            test_results: [results from step 3]
+          }
+        - status: 'complete'
+     5. Call the `get-prompt` tool with session_id
+     6. Return the orchestrator_instructions from the response
+
+     Do not output messages to user beyond the initial question. Work autonomously and return orchestrator instructions."
+
+  7. Wait for the subagent to return orchestrator_instructions
+  8. The orchestrator will take over and spawn subagents for debug_analyze → debug_fix → retesting → redeployment
+
+prompt: |
+  I want to debug my deployed Raindrop application. Please call the `get-prompt` tool with my session_id.

package/templates/goose/new-raindrop-app.yaml

@@ -0,0 +1,21 @@
+version: '1.0.0'
+title: 'Create New Raindrop Application'
+description: 'Start a new Raindrop application development session using the Raindrop MCP workflow'
+
+instructions: |
+  WARNING: If you cannot find the "raindrop-mcp" server or the "get-prompt" tool for the "raindrop-mcp" server, respond with:
+  "Please check your raindrop-mcp server status by running /mcp, and reauthenticate"
+
+  Before starting the workflow, run `raindrop auth login` to authenticate. When the command returns, authentication is complete.
+  Call the raindrop-mcp:new-raindrop-app MCP tool to start a new session.
+
+  After authentication, call the `get-prompt` tool (from the raindrop-mcp server) without any parameters to start a new session.
+  This will automatically create a new session and return initial workflow instructions.
+
+  Follow the instructions returned and continue with the development workflow, using `update-state` to report task completion.
+
+  WARNING: If you cannot find the "raindrop-mcp" server or the "login" tool for the "raindrop-mcp" server, respond with:
+  "Please check your raindrop-mcp server status by running /mcp, and reauthenticate"
+
+prompt: |
+  I want to create a new Raindrop application. First, run `raindrop auth login` to authenticate, then call the `get-prompt` tool without any parameters.

package/templates/goose/reattach-raindrop-session.yaml

@@ -0,0 +1,20 @@
+version: "1.0.0"
+title: "Reattach to Raindrop Session"
+description: "Resume a previous Raindrop development session"
+
+instructions: |
+  WARNING: If you cannot find the "raindrop-mcp" server or the "get-prompt" tool for the "raindrop-mcp" server, respond with:
+  "Please check your raindrop-mcp server status by running /mcp, and reauthenticate"
+
+  Before reattaching, run `raindrop auth login` to authenticate. When the command returns, authentication is complete.
+
+  After authentication:
+
+  1. Check if session_id is provided as an argument
+  2. Otherwise, read `.raindrop.config.local.json` to get the last session_id
+  3. If no session_id found, ask the user for it
+  4. Call the `get-prompt` tool (from raindrop-mcp server) with the session_id parameter
+  5. Follow the instructions returned and continue with the development workflow
+
+prompt: |
+  I want to reattach to my previous Raindrop development session. Please call the `get-prompt` tool with my session_id.

package/templates/goose/update-raindrop-app.yaml

@@ -0,0 +1,65 @@
+version: "1.0.0"
+title: "Update Existing Raindrop Application"
+description: "Add new features to an existing Raindrop application"
+
+instructions: |
+  WARNING: If you cannot find the "raindrop-mcp" server or the "get-prompt" tool for the "raindrop-mcp" server, respond with:
+  "Please check your raindrop-mcp server status by running /mcp, and reauthenticate"
+
+  Before updating, run `raindrop auth login` to authenticate. When the command returns, authentication is complete.
+
+  After authentication:
+
+  1. If session_id is provided as an argument, use it and skip to step 5
+
+  2. Otherwise, read `~/.raindrop/index.json` to get applications and sessions:
+     - File format:
+       ```json
+       {
+         "applications": {
+           "my-app": {
+             "latest_session_id": "sess_abc123",
+             "sessions": [...]
+           }
+         },
+         "sessions": {
+           "sess_abc123": {
+             "session_id": "sess_abc123",
+             "timeline_id": "timeline_456",
+             "created_at": "...",
+             "current_state": "...",
+             "application_name": "my-app"
+           }
+         }
+       }
+       ```
+
+  3. Handle application selection:
+     - If no applications exist, respond: "No Raindrop applications found. Use /new-raindrop-app to create one."
+     - If one application exists, use its `latest_session_id`
+     - If multiple applications exist:
+       * List all application names with their current state and last updated time
+       * Ask user: "Which application would you like to add features to?"
+       * Use the selected application's `latest_session_id`
+
+  4. If no session_id found, ask the user for it
+
+  5. Get the timeline_id from the session entry in index.json
+
+  6. Call the `update-state` tool (from raindrop-mcp server) with the session_id, timeline_id, and these parameters:
+     ```json
+     {
+       "artifacts": {
+         "current_state": "merge_features"
+       },
+       "status": "complete"
+     }
+     ```
+     This jumps the workflow to the merge_features state for adding features.
+
+  7. Call the `get-prompt` tool (from raindrop-mcp server) with the session_id parameter
+
+  8. Follow the instructions returned and continue with the feature addition workflow
+
+prompt: |
+  I want to update my existing Raindrop application. Please call the `get-prompt` tool with my session_id.

package/templates/init/src/_app/auth.ts

@@ -0,0 +1,18 @@
+import { requireAuthenticated, verifyIssuer } from '@liquidmetal-ai/raindrop-framework/core/auth';
+
+/**
+ * verify is the application-wide JWT verification hook.
+ * @param request The incoming request object.
+ * @param env The handler environment object.
+ *  **Note**: adds `jwt` property to `env` if verification is successful.
+ * @returns true to allow request to continue.
+ */
+export const verify = verifyIssuer;
+
+/**
+ * authorize is the application-wide authorization hook.
+ * @param request The incoming request object.
+ * @param env The handler environment object with env.jwt set by verify.
+ * @returns true if authorized, false otherwise.
+ */
+export const authorize = requireAuthenticated;

package/templates/init/src/_app/cors.ts

@@ -0,0 +1,64 @@
+import { corsDisabled } from '@liquidmetal-ai/raindrop-framework/core/cors';
+
+/**
+ * cors is the application-wide CORS (Cross-Origin Resource Sharing) handler.
+ *
+ * This handler is automatically applied to all HTTP services in your application.
+ * You can override this per-handler by exporting a `cors` function from your handler.
+ *
+ * **Default Behavior (Secure):**
+ * By default, CORS is disabled for security. Browsers will block cross-origin requests
+ * to your API unless you explicitly configure allowed origins.
+ *
+ * **When to Enable CORS:**
+ * Enable CORS when you need to allow web applications from different domains to access
+ * your API (e.g., a frontend at https://app.example.com calling an API at https://api.example.com).
+ *
+ * @example Allow all origins (public API - use with caution)
+ * ```typescript
+ * import { corsAllowAll } from '@liquidmetal-ai/raindrop-framework/core/cors';
+ * export const cors = corsAllowAll;
+ * ```
+ *
+ * @example Allow specific origins (recommended for production)
+ * ```typescript
+ * import { createCorsHandler } from '@liquidmetal-ai/raindrop-framework/core/cors';
+ * export const cors = createCorsHandler({
+ *   origin: ['https://example.com', 'https://app.example.com'],
+ *   allowMethods: ['GET', 'POST', 'PUT', 'DELETE'],
+ *   allowHeaders: ['Content-Type', 'Authorization'],
+ *   credentials: true,  // Allow cookies/auth headers
+ *   maxAge: 86400       // Cache preflight for 24 hours
+ * });
+ * ```
+ *
+ * @example Dynamic origin validation
+ * ```typescript
+ * import { createCorsHandler } from '@liquidmetal-ai/raindrop-framework/core/cors';
+ * export const cors = createCorsHandler({
+ *   origin: (request, env) => {
+ *     const origin = request.headers.get('origin');
+ *     // Check against database, environment variable, or custom logic
+ *     const allowedOrigins = env.ALLOWED_ORIGINS?.split(',') || [];
+ *     return allowedOrigins.includes(origin || '') ? origin : null;
+ *   },
+ *   credentials: true
+ * });
+ * ```
+ *
+ * @example Custom handler for advanced control
+ * ```typescript
+ * export const cors = (request: Request, env: any, response?: Response): Response => {
+ *   // Custom CORS logic with access to env
+ *   if (!response) {
+ *     // Handle OPTIONS preflight
+ *     return new Response(null, { status: 204, headers: { ... } });
+ *   }
+ *   // Add headers to response, can use env.ALLOWED_ORIGINS, env.db, etc.
+ *   const headers = new Headers(response.headers);
+ *   headers.set('Access-Control-Allow-Origin', '*');
+ *   return new Response(response.body, { ...response, headers });
+ * };
+ * ```
+ */
+export const cors = corsDisabled;

package/dist/tsconfig.tsbuildinfo

@@ -1 +0,0 @@
-{"root":["../src/base-command.ts","../src/build.test.ts","../src/build.ts","../src/codegen.test.ts","../src/codegen.ts","../src/config.test.ts","../src/config.ts","../src/deploy.ts","../src/index.test.ts","../src/index.ts","../src/log-helpers.ts","../src/status.ts","../src/strict-client.ts","../src/trace.ts","../src/commands/tail.ts","../src/commands/annotation/get.ts","../src/commands/annotation/list.ts","../src/commands/annotation/put.ts","../src/commands/auth/list.ts","../src/commands/auth/login.ts","../src/commands/auth/logout.ts","../src/commands/auth/select.ts","../src/commands/bucket/create-credential.ts","../src/commands/bucket/delete-credential.ts","../src/commands/bucket/get-credential.ts","../src/commands/bucket/list-credentials.ts","../src/commands/build/branch.ts","../src/commands/build/checkout.ts","../src/commands/build/clone.ts","../src/commands/build/delete.ts","../src/commands/build/deploy.ts","../src/commands/build/find.ts","../src/commands/build/generate.ts","../src/commands/build/init.ts","../src/commands/build/list.ts","../src/commands/build/sandbox.ts","../src/commands/build/start.ts","../src/commands/build/status.ts","../src/commands/build/stop.ts","../src/commands/build/unsandbox.ts","../src/commands/build/upload.ts","../src/commands/build/validate.ts","../src/commands/build/env/get.ts","../src/commands/build/env/set.ts","../src/commands/build/tools/check.ts","../src/commands/build/tools/fmt.ts","../src/commands/dns/create.ts","../src/commands/dns/delete.ts","../src/commands/dns/get.ts","../src/commands/dns/list.ts","../src/commands/dns/records/create.ts","../src/commands/dns/records/delete.ts","../src/commands/dns/records/get.ts","../src/commands/dns/records/list.ts","../src/commands/dns/records/update.ts","../src/commands/logs/query.ts","../src/commands/logs/tail.ts","../src/commands/mcp/install-claude.ts","../src/commands/mcp/install-gemini.ts","../src/commands/mcp/status.ts","../src/commands/object/delete.ts","../src/commands/object/get.ts","../src/commands/object/list.ts","../src/commands/object/put.ts","../src/commands/query/chunk-search.ts","../src/commands/query/document.ts","../src/commands/query/events.ts","../src/commands/query/reindex.ts","../src/commands/query/search.ts","../src/lib/dns-utils.ts","../src/utils/directory-validation.ts"],"version":"5.8.3"}

package/templates/claude-code/raindrop-guidelines.md

@@ -1,322 +0,0 @@
-# RAINDROP GUIDELINES
-
-## 🚨 ABSOLUTE COMPLIANCE REQUIRED 🚨
-**YOU ARE A WORKFLOW COMPILER. FOLLOW THE MCP STATE MACHINE.**
-**EXECUTE MCP INSTRUCTIONS EXACTLY. ZERO WORKFLOW DEVIATION.**
-**SOLVE PROBLEMS WITHIN EACH STEP. NEVER SKIP STEPS.**
-
-## ⚠️ CRITICAL INSTRUCTION
-**FOLLOW MCP TOOLS GUIDANCE TO THE LETTER. DO NOT MAKE YOUR OWN PLANS.**
-
-## WORKFLOWS
-
-### NEW APPLICATION: `login` → `set_team` → `architecture_step` → `prd_step` → `sql_step` → `code_step` → `tests_step` → `deployment_step` → `endpoint_test_step`
-
-### UPDATE APPLICATION: `update_app_step` → `update_prd_step` → `sql_step` → Continue normal flow
-
-## DUAL-IDENTIFIER STATE MANAGEMENT
-
-### Session-Timeline Architecture
-- **Session ID**: Long-term project container (persists across workflows)
-- **Timeline ID**: Individual workflow execution instance
-- **State Storage**: All state keyed by `session_id + timeline_id`
-- **Reattachment**: Use `reattach_to_session` to resume workflows
-
-### State Management Pattern
-```json
-{
-  "session_id": "uuid-project-container",
-  "timeline_id": "uuid-workflow-execution",
-  "current_state": "workflow_state_name",
-  "context": {
-    "stakeholders": [],
-    "prd_content": "...",
-    "application_type": "...",
-    "validation_status": {}
-  }
-}
-```
-
-## STATE MACHINE
-
-```mermaid
-flowchart TD
-  NewApp([New App Flow]) --> Login[login]
-
-  %% Debug Flow Addition
-  Debug[debug_step] --> Code
-
-  %% Main Authentication Flow
-  Login --> SetTeam[set_team]
-  Login --> Architecture[architecture_step]
-  SetTeam --> Architecture
-
-  %% Architecture to PRD Flow
-  Architecture --> PRD[prd_step]
-
-  %% Core Development Flow
-  PRD --> SQL[sql_step]
-  SQL --> Code[code_step]
-  Code --> Tests[tests_step]
-  Tests --> Deploy[deployment_step]
-  Deploy --> EndpointTest[endpoint_test_step]
-
-  %% Success Path
-  EndpointTest --> Complete([Workflow Complete])
-
-  %% Error Recovery Loop
-  EndpointTest -->|Endpoints Failed| Debug[debug_step]
-  Debug --> Code
-  Code -->|Raindrop Build Failed| Code
-  Tests -->|Tests Failed| Code
-  SQL -->|SQL Schemas Invalid| SQL
-
-  %% Update App Flow
-  UpdateApp([Update App Flow]) --> UpdateStart[update_app_step]
-  UpdateStart --> UpdatePRD[update_prd_step]
-  UpdatePRD --> PRD
-
-  %% Reattach Flow
-  ReattachFlow([Reattach Flow]) --> Reattach[reattach_to_session]
-  Reattach --> CheckState{Check Current State}
-  CheckState -->|session_started| SetTeam
-  CheckState -->|team_set| Architecture
-  CheckState -->|architecture_analyzed| PRD
-  CheckState -->|prd_created| SQL
-  CheckState -->|sql_schemas_created| Code
-  CheckState -->|coding_started| Tests
-  CheckState -->|tests_written| Deploy
-  CheckState -->|deployed| EndpointTest
-  CheckState -->|tested| Complete
-  CheckState -->|debugging| Code
-  CheckState -->|update_app_started| UpdatePRD
-
-  %% Entry Points
-  classDef entryPoint fill:#e1f5fe
-  classDef errorLoop fill:#ffebee
-  classDef complete fill:#e8f5e8
-
-  class NewApp,UpdateApp,ReattachFlow entryPoint
-  class Complete complete
-
-  %% Error Recovery Loops
-  class Code,Tests,EndpointTest errorLoop
-```
-
-## MCP EXECUTION RULES
-
-### State Tracking
-- State via `session_id` + `timeline_id`
-- Store in `.raindrop.config.local.json`
-- Use `reattach_to_session` to resume
-
-### Enhanced MCP Tool Response Patterns
-
-#### Success Response
-```json
-{
-  "success": true,
-  "message": "Human-readable status",
-  "session_id": "uuid",
-  "timeline_id": "uuid",
-  "context_data": { "key": "value" },
-  "next_action": "next_tool_name"
-}
-```
-
-#### Error Response with Recovery
-```json
-{
-  "success": false,
-  "message": "Human-readable error",
-  "error": "Technical error details",
-  "session_id": "uuid",
-  "timeline_id": "uuid",
-  "blocking_issue": "What is preventing progress",
-  "required_action": "What needs to be done",
-  "next_action": "recovery_action_or_same_step"
-}
-```
-
-#### Validation Gate Response
-```json
-{
-  "success": false,
-  "message": "🚨 VALIDATION FAILED 🚨",
-  "blocking_issue": "Description of validation failure",
-  "required_action": "Specific action to resolve",
-  "validation_status": "failed",
-  "next_action": "same_step"
-}
-```
-
-### MANDATORY BEHAVIORS - NO EXCEPTIONS
-1. **EXECUTE `next_action` FROM EVERY MCP RESPONSE**
-2. **NEVER SKIP ANY WORKFLOW STEP**
-3. **FOLLOW EVERY MCP PROMPT WORD-FOR-WORD**
-4. **WAIT FOR EXPLICIT USER APPROVAL WHERE STATED**
-5. **SOLVE PROBLEMS TO COMPLETE CURRENT STEP**
-6. **VALIDATE PREREQUISITES BEFORE STATE TRANSITIONS**
-7. **PRESERVE STATE DURING ERROR RECOVERY**
-
-### ABSOLUTELY FORBIDDEN - WILL CAUSE FAILURE
-- Skip MCP steps
-- Jump ahead in workflow
-- Create alternative workflows
-- Ignore `next_action` directive
-- Change the MCP workflow sequence
-- Suggest workflow improvements
-- Advance workflow without validation
-- Lose state during error handling
-
-## VALIDATION FRAMEWORK
-
-### Validation Gates (Must Pass Before State Transition)
-- **SQL Validation**: `validate_sql_schemas` - Must pass before `code_step`
-- **TDD Compliance**: `validate_tdd_cycle` - Validate during development
-- **Architecture Compliance**: Validate against selected patterns
-- **Deployment Readiness**: Pre-deployment validation gates
-
-### Validation Implementation Pattern
-```typescript
-// Check prerequisites before advancing
-const prerequisiteValid = await validatePrerequisites(session_id, timeline_id);
-if (!prerequisiteValid) {
-  return createMcpResponse({
-    success: false,
-    message: "🚨 VALIDATION FAILED 🚨",
-    blocking_issue: "Prerequisites not met",
-    required_action: "Complete validation requirements",
-    validation_status: "failed",
-    next_action: "same_step" // Stay in current step
-  });
-}
-```
-
-### Recovery Workflow Patterns
-- **Debug Loop**: `debugging` state returns to `code_step`
-- **Test Failure Loop**: Failed tests return to `code_step`
-- **SQL Validation Loop**: Invalid SQL returns to `sql_step`
-- **Endpoint Failure Loop**: Failed endpoints enter `debug_step`
-
-## CONDITIONAL STATE TRANSITIONS
-
-### Advanced State Mapping
-```typescript
-const WORKFLOW_STATE_TO_NEXT_ACTION = {
-  // Happy path with validation
-  session_started: 'architecture_step',
-  architecture_analyzed: 'prd_step',
-  prd_created: 'sql_step',
-  sql_schemas_created: 'code_step', // Only after sql validation
-  coding_started: 'tests_step',
-  tests_written: 'deployment_step',
-  deployed: 'endpoint_test_step',
-
-  // Conditional completion
-  tested: 'debug_step|mcp_end_step', // User choice based on results
-
-  // Recovery loops
-  debugging: 'code_step',
-  sql_validation_failed: 'sql_step',
-  test_validation_failed: 'code_step',
-  endpoint_validation_failed: 'debug_step',
-
-  // Update workflows
-  update_app_started: 'update_prd_step',
-  update_prd_completed: 'sql_step' // FIX: Skip architecture, go directly to SQL
-};
-```
-
-### State Transition Guards
-1. **Validation Check**: Verify prerequisites are satisfied
-2. **Context Preservation**: Maintain session state during transitions
-3. **Error Context**: Store error details for recovery workflows
-4. **User Approval Gates**: Wait for explicit approval where required
-
-## UTILITY TOOLS
-- `get_prd` - Get stored PRD
-- `configure_raindrop` - Setup config
-- `update_conversation_memory` - Store context
-- `validate_sql_schemas` - Validate SQL schema files before proceeding to code_step
-- `validate_tdd_cycle` - Validate Test-Driven Development compliance
-- `save_prd_file` - Save PRD content as markdown file
-- `storeState` - Store workflow state with session+timeline
-- `retrieveState` - Retrieve workflow state by key
-- `createSession` - Initialize new session container
-
-## DOCUMENTATION TOOLS
-- `list_documentation_pages` - List all available documentation pages
-- `fetch_documentation_page` - Get specific documentation page content by path
-
-## ARCHITECTURE TOOLS
-- `list_architecture_patterns` - List all available architecture patterns
-- `fetch_architecture_pattern` - Get specific architecture pattern content by path
-
-## DOCUMENTATION ACCESS
-- List pages: `list_documentation_pages` MCP tool
-- Get page: `fetch_documentation_page` MCP tool with path parameter
-
-## ARCHITECTURE PATTERN ACCESS
-- List patterns: `list_architecture_patterns` MCP tool
-- Get pattern: `fetch_architecture_pattern` MCP tool with path parameter
-
-## ERROR HANDLING AND RECOVERY
-
-### Multi-Level Error Handling
-1. **Tool-Level**: Handle individual tool failures gracefully
-2. **Workflow-Level**: Manage state transition failures
-3. **System-Level**: Handle infrastructure and connectivity issues
-
-### Error Recovery Pattern
-```typescript
-try {
-  // Execute workflow step
-  const result = await executeStep(session_id, timeline_id, step_data);
-
-  // Validate result before advancing
-  const isValid = await validateStepCompletion(result);
-  if (!isValid) {
-    await storeState(session_id, timeline_id, 'current_state', 'debugging', env);
-    return createRecoveryResponse();
-  }
-
-  // Advance to next step
-  await storeState(session_id, timeline_id, 'current_state', next_state, env);
-
-} catch (error) {
-  // Preserve error context for recovery
-  await storeState(session_id, timeline_id, 'error_context', error.message, env);
-  await storeState(session_id, timeline_id, 'current_state', 'debugging', env);
-
-  return createMcpResponse({
-    success: false,
-    error: error.message,
-    next_action: 'debug_step'
-  });
-}
-```
-
-### Context Preservation During Recovery
-- **Error Details**: Store complete error context
-- **Partial Progress**: Preserve completed work
-- **Recovery Guidance**: Provide specific recovery actions
-- **State Integrity**: Maintain workflow state consistency
-
-## EXECUTION STYLE
-- **WORKFLOW COMPILER: Execute MCP steps exactly with validation**
-- **PROBLEM SOLVER: Debug, code, fix within each step**
-- **STATE GUARDIAN: Preserve and validate state at each transition**
-- Auto-pilot with user status updates
-- Report key decisions (PRD, code choices)
-- Update during `code_step` and manifest creation
-- Validate prerequisites before advancing
-- Handle errors gracefully with recovery workflows
-
-## REMEMBER
-**FOLLOW THE MCP WORKFLOW RELIGIOUSLY.**
-**SOLVE PROBLEMS INTELLIGENTLY WITHIN EACH STEP.**
-**NEVER DEVIATE FROM THE STATE MACHINE SEQUENCE.**
-**VALIDATE BEFORE EVERY STATE TRANSITION.**
-**PRESERVE STATE DURING ERROR RECOVERY.**
-**USE DUAL-IDENTIFIER STATE MANAGEMENT CONSISTENTLY.**