@pennyfarthing/cyclist 9.3.0 → 10.0.0

package/dist/main.js

@@ -8,13 +8,12 @@
  * This module exports testable functions and constants for unit testing,
  * while the Electron-specific runtime code only executes in Electron context.
  */
-import { createServer as createHttpServer } from 'http';
 import { fileURLToPath } from 'url';
 import { dirname, join, basename } from 'path';
 import { getCurrentPersona, detectPennyfarthingProject, watchAgentChanges } from './pennyfarthing.js';
-import { getStoryInfo, getAllReposGitInfoAsync, writePortFile, cleanupPortFile, writePidFile, cleanupPidFile, readPidFile, isProcessRunning, getOtelConfig, writeApprovalPortFile, cleanupApprovalPortFile } from './server.js';
+import { getStoryInfo, getAllReposGitInfoAsync, writePortFile, cleanupPortFile, writePidFile, cleanupPidFile, readPidFile, isProcessRunning, getOtelConfig } from './server.js';
 import { parseToolStats, createEmptyStats } from './tool-stats.js';
-import { getTokenStats, setTokenStatsCallback, setToolEventCallback, resetTokenStats, resetEventStore, getToolEventsFiltered, getToolTypes, exportAuditLogAsJSON, exportAuditLogAsCSV, getAuditLogStats, getUserEmail, setUserEmailCallback, setBackgroundTaskCallback, setBackgroundTaskStartCallback, trackBackgroundTask, getBackgroundTaskByToolId, getBackgroundTasks, } from './otlp-receiver.js';
+import { getTokenStats, setTokenStatsCallback, setToolEventCallback, resetTokenStats, resetEventStore, getToolEventsFiltered, getToolTypes, exportAuditLogAsJSON, exportAuditLogAsCSV, getAuditLogStats, getUserEmail, setUserEmailCallback, setBackgroundTaskCallback, setBackgroundTaskStartCallback, trackBackgroundTask, completeBackgroundTask, getBackgroundTaskByToolId, getBackgroundTasks, } from './otlp-receiver.js';
 import { ClaudeService } from './claude-service.js';
 import { selectContextTier, getPrimeContextJson } from './prime.js';
 import { isTodoWriteMessage, extractTodos } from './todos.js';
@@ -27,10 +26,8 @@ import { getVerboseMode, setVerboseMode } from './settings-store.js';
 import { parse as parseYaml, stringify as stringifyYaml } from 'yaml';
 import { getCurrentSettings, saveUserSettings, initializeSettings, loadGrants, saveGrants, } from './settings.js';
 import { broadcastBackgroundTaskEvent } from './api/background-tasks.js';
-import { setStoryUpdateCallback, setGitUpdateCallback, broadcastClaudeMessage, setClaudeSendCallback, setClaudeAbortCallback, setClaudeClearCallback, setClaudeSetModeCallback, setClaudeGetModeCallback, setClaudeClearAndReloadCallback, broadcastTodosUpdate, broadcastContextUpdate } from './websocket.js';
-import { initializeGrants, setGrantsPersistCallback } from './settings-store.js';
-// Story 33-7: Import approval gate functions for tool execution pipeline
-import { interceptToolUse, requestApproval, createRejectionError, } from './approval-gate.js';
+import { setStoryUpdateCallback, setGitUpdateCallback, broadcastClaudeMessage, setClaudeSendCallback, setClaudeAbortCallback, setClaudeClearCallback, setClaudeSetModeCallback, setClaudeGetModeCallback, setClaudeClearAndReloadCallback, broadcastTodosUpdate, broadcastContextUpdate, broadcastPanelToggle } from './websocket.js';
+import { initializeGrants, setGrantsPersistCallback, clearSessionGrants } from './settings-store.js';
 import { openSettingsWindow, setMainWindowRef, setBrowserWindowRef } from './settings-window.js';
 import { setBellMode } from './bell-mode.js';
 import { IPC_DATA_CHANNELS, IPC_CLAUDE_CHANNELS, IPC_AGENT_CHANNELS, IPC_SETTINGS_CHANNELS, IPC_AUDIT_LOG_CHANNELS, IPC_FILE_BROWSER_CHANNELS, IPC_COMMAND_CHANNELS, IPC_BACKGROUND_TASK_CHANNELS, IPC_SKILL_CHANNELS, IPC_LAYOUT_CHANNELS, IPC_AVATAR_CHANNELS, } from './ipc-channels.js';
@@ -65,8 +62,9 @@ catch {
 // Re-export IPC channels from dedicated module
 export { IPC_DATA_CHANNELS, IPC_CLAUDE_CHANNELS, IPC_AGENT_CHANNELS, IPC_DIFF_CHANNELS, IPC_SETTINGS_CHANNELS, IPC_AUDIT_LOG_CHANNELS, IPC_FILE_BROWSER_CHANNELS, IPC_COMMAND_CHANNELS, IPC_BACKGROUND_TASK_CHANNELS, IPC_SKILL_CHANNELS, IPC_CONTEXT_CLEAR_CHANNELS, IPC_LAYOUT_CHANNELS, IPC_AVATAR_CHANNELS, } from './ipc-channels.js';
 // Re-export menu builders from dedicated module
-export { AGENT_DEFINITIONS, WORKFLOW_DEFINITIONS, buildAgentMenu, buildWorkflowMenu, buildToolsMenu, buildViewMenu, getMenuTemplate, } from './menu-builder.js';
-import { buildAgentMenu, buildWorkflowMenu, buildToolsMenu, buildViewMenu, } from './menu-builder.js';
+export { AGENT_DEFINITIONS, WORKFLOW_DEFINITIONS, buildAgentMenu, buildWorkflowMenu, buildToolsMenu, buildViewMenu, getMenuTemplate, setPanelToggleBroadcast, } from './menu-builder.js';
+// Local imports for menu building
+import { buildAgentMenu, buildWorkflowMenu, buildToolsMenu, buildViewMenu, setPanelToggleBroadcast, } from './menu-builder.js';
 /**
  * Get list of registered data IPC channels (for testing)
  * Returns the data channels that setupDataIPCHandlers will register
@@ -794,6 +792,31 @@ export function startProjectWatchers() {
                 // MSSCI-14190: Process tool_use messages for diff tracking (same as IPC handler)
                 // This was missing from the WebSocket callback path!
                 processToolUseFromMessage(message);
+                // Complete subagent tasks when tool_result arrives
+                // CLI format: discrete tool_result message
+                if (message.type === 'tool_result') {
+                    const msg = message;
+                    if (msg.tool_id) {
+                        const task = getBackgroundTaskByToolId(msg.tool_id);
+                        if (task) {
+                            completeBackgroundTask(msg.tool_id, !msg.is_error, msg.is_error ? undefined : msg.output?.slice(0, 500), msg.is_error ? (msg.output?.slice(0, 500) || 'Task failed') : undefined);
+                        }
+                    }
+                }
+                // SDK format: tool_result nested in user message content
+                if (message.type === 'user') {
+                    const userMsg = message;
+                    if (userMsg.message?.content) {
+                        for (const block of userMsg.message.content) {
+                            if (block.type === 'tool_result' && block.tool_use_id) {
+                                const task = getBackgroundTaskByToolId(block.tool_use_id);
+                                if (task) {
+                                    completeBackgroundTask(block.tool_use_id, !block.is_error, block.is_error ? undefined : (typeof block.content === 'string' ? block.content.slice(0, 500) : undefined), block.is_error ? (typeof block.content === 'string' ? block.content.slice(0, 500) : 'Task failed') : undefined);
+                                }
+                            }
+                        }
+                    }
+                }
             }
             onComplete();
             broadcastToRenderer(IPC_CLAUDE_CHANNELS.CLAUDE_COMPLETE, null);
@@ -951,16 +974,18 @@ function processToolUseFromMessage(message) {
         // Store for OTEL correlation
         if (toolId && toolInput) {
             storePendingToolInput(toolId, toolName, toolInput);
-            // MSSCI-14210: Track background Task tools
-            if (toolName === 'Task' && toolInput.run_in_background === true) {
-                const description = toolInput.description || toolInput.prompt?.substring(0, 50) || 'Background task';
+            // Track all Task tool subagents (background and foreground)
+            // All must be tracked so enrichMessageWithSubagentContext can look them up
+            if (toolName === 'Task') {
+                const description = toolInput.description || toolInput.prompt?.substring(0, 50) || 'Subagent task';
                 const subagentType = toolInput.subagent_type || 'general-purpose';
+                const isBackground = toolInput.run_in_background === true;
                 trackBackgroundTask({
                     taskId: toolId,
                     description,
                     subagentType,
                     startedAt: Date.now(),
-                    isBackground: true,
+                    isBackground,
                 });
             }
         }
@@ -1506,363 +1531,6 @@ export function setupCommandIPCHandlers(ipcMain) {
     registeredCommandChannels = [IPC_COMMAND_CHANNELS.EXECUTE];
     console.log('Command IPC handlers registered');
 }
-// Dependency injection for testing
-let ipcSender = null;
-let toolExecutor = null;
-let errorInjector = null;
-/**
- * Set the IPC sender function (for testing)
- */
-export function setIPCSender(sender) {
-    ipcSender = sender;
-}
-/**
- * Set the tool executor function (for testing)
- */
-export function setToolExecutor(executor) {
-    toolExecutor = executor;
-}
-/**
- * Set the error injector function (for testing)
- */
-export function setErrorInjector(injector) {
-    errorInjector = injector;
-}
-/**
- * Send an approval request to the renderer via IPC
- */
-export function sendApprovalRequest(toolId, toolName, context) {
-    const sender = ipcSender || broadcastToRenderer;
-    sender('permission-request', {
-        toolId,
-        toolName,
-        context,
-    });
-}
-/**
- * Handle permission response from renderer
- * Called by IPC handler when user responds to approval modal
- */
-export function handlePermissionResponse(response) {
-    // Story 33-7: First try to resolve hook approval (from PreToolUse hook)
-    // This is the path that actually controls tool execution
-    resolveHookApproval(response.toolId, response.approved, response.grantScope);
-    // Also resolve approval-gate.js pending approvals for backwards compatibility
-    // (This was the old observer-only path)
-    import('./approval-gate.js').then(({ resolveApproval }) => {
-        resolveApproval(response.toolId, response.approved, response.grantScope);
-    });
-}
-/**
- * Process a tool_use message with approval gate check
- * This is the main integration point for story 33-7
- *
- * @param message - The tool_use message to process
- * @returns ApprovalResult indicating whether approval is needed and outcome
- */
-export async function processToolUseWithApproval(message) {
-    // Check if this tool_use needs approval
-    const interceptResult = interceptToolUse(message);
-    // If gate is disabled or grant exists, pass through immediately
-    if (!interceptResult.shouldApprove) {
-        // Execute tool if executor is set
-        if (toolExecutor) {
-            toolExecutor(message);
-        }
-        return {
-            needsApproval: false,
-            passThrough: true,
-        };
-    }
-    // Need approval - send IPC request and wait for response
-    sendApprovalRequest(interceptResult.toolId, interceptResult.toolName, interceptResult.context);
-    // Get the command for Bash tools, or use context for other tools
-    const command = interceptResult.toolName === 'Bash'
-        ? interceptResult.context.command || ''
-        : JSON.stringify(interceptResult.context);
-    // Wait for user response
-    const approved = await requestApproval(command, interceptResult.toolId);
-    if (approved) {
-        // User approved - execute tool
-        if (toolExecutor) {
-            toolExecutor(message);
-        }
-        return {
-            needsApproval: true,
-            passThrough: true,
-            approved: true,
-        };
-    }
-    else {
-        // User rejected - create and inject error
-        const errorMessage = createRejectionError(interceptResult.toolId);
-        if (errorInjector) {
-            errorInjector(errorMessage);
-        }
-        return {
-            needsApproval: true,
-            passThrough: false,
-            approved: false,
-            rejected: true,
-            errorMessage,
-        };
-    }
-}
-/**
- * Set up IPC handlers for approval gate
- * Story 33-7: Handles permission request/response flow
- */
-export function setupApprovalIPCHandlers(ipcMain) {
-    // Handle permission response from renderer
-    if (ipcMain.on) {
-        ipcMain.on('permission-response', (_event, response) => {
-            handlePermissionResponse(response);
-        });
-    }
-    if (ipcMain.handle) {
-        ipcMain.handle('permission-response', async (_event, response) => {
-            handlePermissionResponse(response);
-            return { success: true };
-        });
-    }
-    console.log('Approval gate IPC handlers registered');
-}
-// =============================================================================
-// Story 33-7: Approval Hook Server
-// =============================================================================
-// HTTP server that receives approval requests from the PreToolUse hook script.
-// The hook runs in Claude Code's process, sends requests here, we show modal,
-// user decides, we respond, hook tells Claude Code to allow/deny.
-//
-// Multi-instance support: Uses dynamic port selection with .cyclist-approval-port
-// discovery file to prevent cross-instance interference when multiple Cyclist
-// windows are open for different projects.
-let approvalServer = null;
-let approvalServerPort = null;
-// Pending approval requests from hooks, keyed by toolId
-// MSSCI-11947: Extended to support data field for interactive tools
-const pendingHookApprovals = new Map();
-/**
- * Handle incoming approval request from hook script
- */
-async function handleHookApprovalRequest(toolName, toolId, input) {
-    // Check if gate is enabled
-    const { getBashApprovalGate, checkGrant, isAllowlisted } = await import('./settings-store.js');
-    if (!getBashApprovalGate()) {
-        return { decision: 'allow', reason: 'Approval gate disabled' };
-    }
-    // Check allowlist and grants
-    if (toolName === 'Bash') {
-        const command = input.command || '';
-        if (isAllowlisted(command) || checkGrant('Bash', command)) {
-            return { decision: 'allow', reason: 'Matched allowlist or existing grant' };
-        }
-    }
-    // Need user approval - send to renderer and wait
-    return new Promise((resolve) => {
-        pendingHookApprovals.set(toolId, { resolve, toolName, input });
-        // Send approval request to renderer
-        broadcastToRenderer('permission-request', {
-            toolId,
-            toolName,
-            context: input,
-            source: 'hook', // Indicate this came from hook, not observation
-        });
-    });
-}
-/**
- * Resolve a pending hook approval (called when user responds to modal)
- */
-export function resolveHookApproval(toolId, approved, grantScope) {
-    const pending = pendingHookApprovals.get(toolId);
-    if (pending) {
-        // Add grant if approved with scope
-        if (approved && grantScope && pending.toolName === 'Bash') {
-            import('./settings-store.js').then(({ addGrant, extractPattern }) => {
-                const command = pending.input.command || '';
-                const pattern = extractPattern(command);
-                addGrant({
-                    tool: 'Bash',
-                    scope: pattern,
-                    grant_type: grantScope,
-                    granted_at: new Date().toISOString(),
-                });
-            });
-        }
-        pending.resolve({
-            decision: approved ? 'allow' : 'deny',
-            reason: approved ? `Approved by user (${grantScope || 'once'})` : 'Rejected by user',
-        });
-        pendingHookApprovals.delete(toolId);
-    }
-}
-// =============================================================================
-// MSSCI-11947: Hook Response Data Channel
-// =============================================================================
-// Functions for handling interactive tools (AskUserQuestion, ExitPlanMode)
-// that need to return structured data back to Claude, not just allow/deny.
-/**
- * Check if a tool_use is an interactive tool that needs data return
- * MSSCI-11947: AC1 - Detect AskUserQuestion and ExitPlanMode
- */
-export function isInteractiveToolUse(toolUse) {
-    const interactiveTools = ['AskUserQuestion', 'ExitPlanMode'];
-    return toolUse.type === 'tool_use' && interactiveTools.includes(toolUse.tool_name || '');
-}
-/**
- * Process an interactive tool_use and wait for user response with data
- * MSSCI-11947: AC1 - Handle interactive tool approval flow
- */
-export function processInteractiveToolUse(toolUse) {
-    const toolId = toolUse.tool_id || `interactive-${Date.now()}`;
-    const toolName = toolUse.tool_name || 'Unknown';
-    const input = toolUse.input || {};
-    return new Promise((resolve) => {
-        pendingHookApprovals.set(toolId, { resolve, toolName, input });
-        // Send approval request to renderer with full tool input
-        broadcastToRenderer('permission-request', {
-            toolId,
-            toolName,
-            context: input,
-            source: 'hook',
-        });
-    });
-}
-/**
- * Resolve a pending hook approval with data (for interactive tools)
- * MSSCI-11947: AC1 - Extended resolve that includes data field
- */
-export function resolveHookApprovalWithData(toolId, approved, grantScope, data) {
-    const pending = pendingHookApprovals.get(toolId);
-    if (pending) {
-        // For interactive tools, we don't create grants (they're one-time responses)
-        pending.resolve({
-            decision: approved ? 'allow' : 'deny',
-            reason: approved ? `Approved by user (${grantScope || 'once'})` : 'Rejected by user',
-            data: data || {},
-        });
-        pendingHookApprovals.delete(toolId);
-    }
-}
-/**
- * Format hook response with optional data field
- * MSSCI-11947: AC4 - Format response for hook output
- */
-export function formatHookResponseWithData(decision, reason, data) {
-    const response = {
-        decision,
-        reason,
-    };
-    if (data !== undefined) {
-        response.data = data;
-    }
-    return response;
-}
-/**
- * Format answers as updatedInput for AskUserQuestion
- * MSSCI-11947: AC4 - Format for hook updatedInput
- */
-export function formatUpdatedInputForAskUserQuestion(answers) {
-    return { answers };
-}
-/**
- * Format plan response as updatedInput for ExitPlanMode
- * MSSCI-11947: AC4 - Format for hook updatedInput
- */
-export function formatUpdatedInputForExitPlanMode(response) {
-    return response;
-}
-/**
- * Serialize approval data for transmission
- * MSSCI-11947: AC1 - JSON serialization helper
- */
-export function serializeApprovalData(data) {
-    return JSON.stringify(data);
-}
-/**
- * Deserialize approval data from transmission
- * MSSCI-11947: AC1 - JSON deserialization helper
- */
-export function deserializeApprovalData(data) {
-    return JSON.parse(data);
-}
-/**
- * Start the approval hook server with dynamic port selection
- * Uses port 0 to let OS assign an available port (avoids race conditions)
- * Writes port to .cyclist-approval-port for hook discovery
- */
-export async function startApprovalServer() {
-    if (approvalServer) {
-        console.log('Approval server already running');
-        return;
-    }
-    const projectDir = getProjectDirectory();
-    if (!projectDir) {
-        console.warn('No project directory set, cannot start approval server');
-        return;
-    }
-    approvalServer = createHttpServer(async (req, res) => {
-        if (req.method === 'POST' && req.url === '/approval-request') {
-            let body = '';
-            req.on('data', (chunk) => { body += chunk; });
-            req.on('end', async () => {
-                try {
-                    const data = JSON.parse(body);
-                    const { toolName, toolId, input } = data;
-                    const response = await handleHookApprovalRequest(toolName, toolId, input);
-                    res.writeHead(200, { 'Content-Type': 'application/json' });
-                    res.end(JSON.stringify(response));
-                }
-                catch {
-                    res.writeHead(400, { 'Content-Type': 'application/json' });
-                    res.end(JSON.stringify({ error: 'Invalid request' }));
-                }
-            });
-        }
-        else {
-            res.writeHead(404);
-            res.end('Not found');
-        }
-    });
-    // Use port 0 to let OS assign an available port (avoids race conditions)
-    approvalServer.listen(0, '127.0.0.1', () => {
-        const addr = approvalServer.address();
-        approvalServerPort = typeof addr === 'object' && addr ? addr.port : null;
-        if (approvalServerPort) {
-            console.log(`Approval hook server running on http://127.0.0.1:${approvalServerPort}`);
-            // Write port file for hook discovery
-            writeApprovalPortFile(projectDir, approvalServerPort);
-            console.log(`[33-7] Wrote .cyclist-approval-port file to ${projectDir}`);
-        }
-    });
-    approvalServer.on('error', (err) => {
-        console.error('Approval server error:', err);
-        approvalServerPort = null;
-    });
-}
-/**
- * Stop the approval hook server and clean up port file
- */
-export function stopApprovalServer() {
-    if (approvalServer) {
-        approvalServer.close();
-        approvalServer = null;
-        approvalServerPort = null;
-        // Clean up port file
-        const projectDir = getProjectDirectory();
-        if (projectDir) {
-            cleanupApprovalPortFile(projectDir);
-            console.log('[33-7] Cleaned up .cyclist-approval-port file');
-        }
-        console.log('Approval hook server stopped');
-    }
-}
-/**
- * Get the current approval server port (for testing)
- */
-export function getApprovalServerPort() {
-    return approvalServerPort;
-}
 // =============================================================================
 // Session Persistence (E7-3: AC4)
 // =============================================================================
@@ -2074,8 +1742,6 @@ if (isElectron) {
     setupAuditLogIPCHandlers(ipcMain);
     setupCommandIPCHandlers(ipcMain); // 23-3: Command execution
     setupSkillIPCHandlers(ipcMain); // 35-12: Skill invocation tracking
-    setupApprovalIPCHandlers(ipcMain); // 33-7: Approval gate wiring
-    // NOTE: startApprovalServer() moved to app.whenReady() - needs project directory
     /**
      * Kill orphaned Claude CLI process from previous Cyclist session in THIS project.
      * B-24 fix: Only kills the specific PID from .cyclist-pid, not all Claude processes.
@@ -2225,12 +1891,13 @@ if (isElectron) {
             // In Electron mode, Claude messages are broadcast from main.ts, not per-connection ClaudeService
             process.env.CYCLIST_ELECTRON_MODE = '1';
             await startServer();
-            await startApprovalServer(); // 33-7: Start after project dir set
             createWindow();
             if (mainWindow && projectDir) {
                 mainWindow.setTitle(`Cyclist - ${basename(projectDir)}`);
             }
             // B-23: Wire agent and workflow menus to Electron menu bar
+            // Wire panel toggle to WebSocket broadcast
+            setPanelToggleBroadcast(broadcastPanelToggle);
             // Use standard macOS menu roles instead of reconstructing existing menu
             // (reconstructing fails on nested submenus like Window)
             // 22-5: Custom View menu with Verbose Mode toggle
@@ -2298,6 +1965,8 @@ if (isElectron) {
         if (claudeServiceInstance) {
             claudeServiceInstance.abort();
         }
+        // MSSCI-14324: Clear session/once grants on shutdown
+        clearSessionGrants();
         // B-24 fix: Clean up PID file on graceful shutdown
         const projectDir = getProjectDirectory();
         if (projectDir) {