@adhdev/daemon-core 0.5.38 → 0.5.41

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@adhdev/daemon-core",
-  "version": "0.5.38",
+  "version": "0.5.41",
   "description": "ADHDev daemon core — CDP, IDE detection, providers, command execution",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/providers/_builtin/cli/claude-cli/provider.json

@@ -86,7 +86,7 @@
       { "source": "Always\\s*allow", "flags": "i" },
       { "source": "\\(y/n\\)", "flags": "i" },
       { "source": "\\[Y/n\\]", "flags": "i" },
-      { "source": "Run\\s+\\w+\\s+command", "flags": "i" },
+      { "source": "Run\\s*this\\s*command", "flags": "i" },
       { "source": "Allow\\s*tool", "flags": "i" },
       { "source": "Yes,?\\s*don'?t\\s*ask", "flags": "i" },
       { "source": "Deny", "flags": "i" },

package/providers/_builtin/extension/codex/scripts/1.0/explore_dom.js

@@ -45,17 +45,11 @@
 
     // 6. Check if we're on a task list or chat view
     const hasConversationList = document.querySelectorAll('[role="button"][class*="rounded-lg"]').length > 0;
-    const headerEl = document.querySelector('[style*="view-transition-name: header-title"]');
-    const headerText = (headerEl?.textContent || '').trim();
-    const isTaskList = headerText === '작업' || headerText === 'Tasks';
-
-    if (isTaskList) {
-      return JSON.stringify({ __adhdev_skip_iframe: true, error: 'Found Tasks webview instead of Chat' });
-    }
+    const headerText = document.querySelector('[style*="view-transition-name: header-title"]')?.textContent?.trim() || '';
 
     return JSON.stringify({
       headerText,
-      isTaskList,
+      isTaskList: headerText === '작업' || headerText === 'Tasks',
       modelFound: modelMatch ? modelMatch[0] : null,
       textareaCount: textareas.length,
       textareas: Array.from(textareas).map(t => ({

package/providers/_builtin/extension/codex/scripts/1.0/read_chat.js

@@ -52,11 +52,6 @@
     const headerEl = doc.querySelector('[style*="view-transition-name: header-title"]');
     const headerText = (headerEl?.textContent || '').trim();
     const isTaskList = headerText === '작업' || headerText === 'Tasks';
-    
-    // If we accidentally evaluated inside the Tasks webview instead of Chat, tell Daemon to try the next matching webview
-    if (isTaskList) {
-      return JSON.stringify({ __adhdev_skip_iframe: true, error: 'Found Tasks webview instead of Chat' });
-    }
 
     // ─── Rich content extractor ───
     const BLOCK_TAGS = new Set(['DIV', 'P', 'BR', 'LI', 'TR', 'H1', 'H2', 'H3', 'H4', 'H5', 'H6', 'BLOCKQUOTE', 'HR', 'SECTION', 'ARTICLE']);
@@ -296,145 +291,24 @@
 
     // ─── 3. Status ───
     let status = 'idle';
-    // Filter out disabled buttons to avoid matching old historical prompts
-    // Include roles, inputs, custom vscode tags, and generic interactive lists used by multiple-choice forms
-    const selectors = [
-      'button', '[role="radio"]', '[role="button"]', '[role="option"]', '[role="menuitem"]',
-      'input[type="radio"] + label', 'input[type="radio"] ~ span', 'input[type="checkbox"] + label',
-      'vscode-button', 'vscode-radio', 'vscode-checkbox', 'vscode-option', 'li'
-    ].join(', ');
-    
-    const buttons = Array.from(doc.querySelectorAll(selectors))
-      .filter(b => b.offsetWidth > 0 && !b.disabled && !b.closest('[inert]'));
-    
-    const getBtnLabel = (b) => {
-      let t = (b.textContent || '').trim();
-      return t || (b.getAttribute('aria-label') || '').trim();
-    };
-    
-    const buttonLabels = buttons.map(getBtnLabel).map(t => t.toLowerCase());
-
-    if (buttonLabels.some(l => l.includes('cancel') || l.includes('취소') || l.includes('stop') || l.includes('중지'))) {
+    const buttons = Array.from(doc.querySelectorAll('button'))
+      .filter(b => b.offsetWidth > 0);
+    const buttonTexts = buttons.map(b => (b.textContent || '').trim().toLowerCase());
+    const buttonLabels = buttons.map(b => (b.getAttribute('aria-label') || '').toLowerCase());
+
+    if (buttonTexts.includes('cancel') || buttonTexts.includes('취소') ||
+        buttonLabels.some(l => l.includes('cancel') || l.includes('취소') || l.includes('stop') || l.includes('중지'))) {
       status = 'generating';
     }
 
-    // ─── 5. Universal Approval Modal Detection (Language-Agnostic) ───
-    // The approval/interaction panel in Codex is rendered OUTSIDE the chat scroll area,
-    // typically in a sibling "request-input-panel" region at the bottom.
-    // We must scan the ENTIRE document body to find it.
-    let activeModal = null;
-
-    // Look for the request-input-panel area first (contains radio options + submit/skip)
-    // The approval panel lives in the outer webview document, NOT inside the inner iframe (doc).
-    // Use `document` (the webview frame root) to find it, same as explore_dom.js does.
-    const searchDocs = [document, doc]; // outer first, then inner iframe
-    let requestPanel = null;
-    for (const d of searchDocs) {
-      requestPanel = d.querySelector('[class*="request-input-panel"]');
-      if (requestPanel) break;
-      // Also search all textareas for the class
-      const tas = d.querySelectorAll('textarea');
-      for (const ta of tas) {
-        if (ta.className && ta.className.includes('request-input-panel')) {
-          requestPanel = ta;
-          break;
-        }
-      }
-      if (requestPanel) break;
+    // Codex approval buttons appear as distinct action buttons (e.g. "Approve", "Always approve", "Deny")
+    // They typically appear at the bottom of a tool-call/action block
+    const approvalSpecificPatterns = /^(approve|always approve|deny|reject|승인|항상 승인|거부)/i;
+    const hasApprovalButton = buttonTexts.some(b => approvalSpecificPatterns.test(b)) ||
+        buttonLabels.some(l => approvalSpecificPatterns.test(l));
+    if (hasApprovalButton) {
+      status = 'waiting_approval';
     }
-    // Walk up from the request-input-panel to find the full approval card
-    let approvalArea = null;
-    if (requestPanel) {
-      let p = requestPanel;
-      for (let i = 0; i < 12 && p && p.parentElement; i++) {
-        p = p.parentElement;
-        const btns = p.querySelectorAll('button').length;
-        const radios = p.querySelectorAll('[role="radio"], [role="option"]').length;
-        const total = btns + radios;
-        if (btns >= 4) {
-          // Found the approval card with enough interactive elements
-          approvalArea = p;
-          break;
-        }
-      }
-    }
-
-
-
-
-    if (approvalArea) {
-      // ─── Codex Approval Panel (request-input-panel based) ───
-      // The approval form has:
-      //   - Option items as plain divs (e.g., "1.\n예", "2.\n네,...", "3.\n아니요,...")
-      //   - Action buttons (e.g., "건너뛰기", "제출⏎") as <button> elements
-      // We need to extract BOTH for the dashboard.
-      
-      // Find parent container that has all options (level 3 from textarea, btns >= 4)
-      // approvalArea is already set to this level.
-      
-      // Get the prompt message (from grandparent that starts with "Do you want...")
-      let messageText = '';
-      let msgParent = approvalArea.parentElement;
-      for (let i = 0; i < 5 && msgParent; i++) {
-        const t = (msgParent.innerText || '').trim();
-        if (t.length > 20 && /[?？]/.test(t.substring(0, 200))) {
-          // Found a parent whose text starts with a question
-          messageText = t.split('\n')[0].trim();
-          break;
-        }
-        msgParent = msgParent.parentElement;
-      }
-      if (!messageText) {
-        // Fallback: use the approvalArea parent's text, first sentence
-        const parentText = (approvalArea.parentElement?.innerText || approvalArea.innerText || '').trim();
-        const firstLine = parentText.split('\n')[0].trim();
-        messageText = firstLine.length > 5 ? firstLine : 'Agent requires an interaction';
-      }
-
-      // Extract option labels from the panel's direct/nested children
-      const allBtns = Array.from(approvalArea.querySelectorAll('button'))
-        .filter(b => b.offsetWidth > 0 && !b.disabled)
-        .map(b => (b.textContent || '').trim())
-        .filter(t => t.length > 0 && t.length < 150);
-
-      // Extract numbered option text items (the entire panel text, split by option numbering)
-      const panelText = (approvalArea.innerText || '').trim();
-      // Parse "1.\n예\n2.\n네,...\n3.\n아니요,...\n건너뛰기\n제출⏎" format
-      const optionMatches = panelText.match(/\d+\.\n[^\n]+(?:\n[^\d][^\n]*)*/g) || [];
-      let options = optionMatches.map(o => o.replace(/\n/g, '').trim()).filter(o => o.length > 0);
-      
-      // Clean: remove any trailing button labels that got captured in the last option
-      const btnLabelsSet = new Set(allBtns);
-      options = options.map(opt => {
-        let changed = true;
-        while (changed) {
-          changed = false;
-          for (const bl of btnLabelsSet) {
-            if (opt.endsWith(bl)) {
-              opt = opt.slice(0, -bl.length).trim();
-              changed = true;
-            }
-          }
-        }
-        return opt;
-      }).filter(o => o.length > 0);
-
-      // Merge numbered options from both sources (buttons + text parsing)
-      const allNumbered = [...new Set([...allBtns.filter(b => /^\d+\./.test(b)), ...options])];
-      allNumbered.sort((a, b) => (parseInt(a) || 999) - (parseInt(b) || 999));
-      const actionBtns = allBtns.filter(b => !/^\d+\./.test(b));
-
-      // Combine: sorted numbered options first, then action buttons (deduped)
-      const uniqueActions = [...new Set([...allNumbered, ...actionBtns])];
-
-      if (uniqueActions.length > 0) {
-        status = 'waiting_approval';
-        activeModal = {
-          message: messageText,
-          buttons: uniqueActions
-        };
-      }
-    } // end if (approvalArea)
 
     if (isTaskList) {
       status = messages.length === 0 ? 'idle' : status;
@@ -442,53 +316,30 @@
     if (!isVisible && messages.length === 0) status = 'panel_hidden';
 
     // ─── 4. Model / Mode ───
-    // Language-agnostic detection via DOM structure.
-    // Model: button text matching model name patterns (GPT-*, o1-*, claude-* — always English).
-    // Mode: non-model aria-haspopup="menu" button in composer area (language-agnostic).
     let model = '';
     let mode = '';
-    for (const d of [doc, document]) {
-      // Search in composer area, or common footer containers
-      const searchRoots = [
-        d.querySelector('[class*="thread-composer-max-width"]'),
-        d.querySelector('[class*="thread-composer"]'),
-        d.querySelector('[class*="pb-2"]'),
-        d.body,
-      ].filter(Boolean);
-
-      for (const searchRoot of searchRoots) {
-        if (model && mode) break;
-
-        // aria-haspopup="menu" buttons — dropdown triggers for model/mode
-        if (!model || !mode) {
-          const menuBtns = Array.from(searchRoot.querySelectorAll('button[aria-haspopup="menu"]'))
-            .filter(b => b.offsetWidth > 0);
-          for (const btn of menuBtns) {
-            const text = (btn.textContent || '').trim();
-            if (!model && /^(GPT-|gpt-|o\d|claude-|sonnet|opus)/i.test(text)) {
-              model = text;
-            } else if (!mode && text.length > 0 && text.length < 30) {
-              mode = text;
-            }
-          }
-        }
-
-        // Fallback: any visible button with model-like text
-        if (!model) {
-          const allBtns = Array.from(searchRoot.querySelectorAll('button'))
-            .filter(b => b.offsetWidth > 0);
-          for (const btn of allBtns) {
-            const text = (btn.textContent || '').trim();
-            if (/^(GPT-|gpt-|o\d|claude-|sonnet|opus)/i.test(text)) {
-              model = text;
-              break;
-            }
-          }
-        }
+    const footerButtons = doc.querySelectorAll(
+      '[class*="thread-composer-max-width"] button, [class*="pb-2"] button'
+    );
+    for (const btn of footerButtons) {
+      const text = (btn.textContent || '').trim();
+      if (/^(GPT-|gpt-|o\d|claude-)/i.test(text)) model = text;
+      if (/^(낮음|중간|높음|low|medium|high)$/i.test(text)) mode = text;
+    }
 
-        if (model) break; // found model, no need to keep searching this scope
+    // ─── 5. Approval modal ───
+    let activeModal = null;
+    if (status === 'waiting_approval') {
+      const approvalBtns = buttons
+        .map(b => (b.textContent || '').trim())
+        .filter(t => t && t.length > 0 && t.length < 40 &&
+          /approve|accept|allow|confirm|run|proceed|cancel|deny|reject|승인|허용|실행|취소|거부/i.test(t));
+      if (approvalBtns.length > 0) {
+        activeModal = {
+          message: 'Codex wants to perform an action',
+          buttons: [...new Set(approvalBtns)],
+        };
       }
-      if (model) break; // found in this frame
     }
 
     // ─── 6. Task info ───

package/providers/_builtin/extension/codex/scripts/1.0/resolve_action.js

@@ -2,79 +2,39 @@
  * Codex Extension — resolve_action
  *
  * Clicks approval/denial buttons in the Codex UI.
- * Actions: "approve", "deny", "cancel", or raw button text
+ * Actions: "approve", "deny", "cancel"
  *
- * Placeholders: ${action}, ${button}
+ * Placeholder: ${ACTION}
  */
 (() => {
   try {
-    const action = ${action};
-    const buttonText = ${button};
+    const action = ${ACTION};
     
-    // Search in document (outer webview frame where Codex approval panel lives)
-    const buttons = Array.from(document.querySelectorAll('button, [role="radio"], [role="button"], input[type="radio"] + label'))
-      .filter(b => b.offsetWidth > 0 && !b.disabled && !b.closest('[inert]'));
+    const buttons = Array.from(document.querySelectorAll('button'))
+      .filter(b => b.offsetWidth > 0);
 
-    const actionLower = (action || '').toLowerCase();
+    // Map action to button text patterns
     const patterns = {
       approve: /^(approve|accept|allow|confirm|run|proceed|yes|승인|허용|실행|확인)/i,
       deny: /^(deny|reject|no|거부|아니오)/i,
       cancel: /^(cancel|stop|취소|중지)/i,
     };
 
-    // Determine what to search for: use buttonText if provided, otherwise match action pattern
-    const searchText = buttonText || action || '';
-    let targetBtn = null;
+    const pattern = patterns[action.toLowerCase()] || patterns.approve;
 
-    const getBtnLabel = (b) => {
-      let t = (b.textContent || '').trim();
-      return t || (b.getAttribute('aria-label') || '').trim();
-    };
-
-    // 1. Try exact match on buttonText first
-    if (buttonText) {
-      targetBtn = buttons.find(b => getBtnLabel(b) === buttonText);
-      // 2. Try startsWith match (for cases where button text has extra chars like ⏎)
-      if (!targetBtn) {
-        targetBtn = buttons.find(b => getBtnLabel(b).startsWith(buttonText));
-      }
-      // 3. Try includes match  
-      if (!targetBtn) {
-        targetBtn = buttons.find(b => getBtnLabel(b).includes(buttonText));
-      }
-    }
-
-    // 4. Fall back to pattern-based matching
-    if (!targetBtn && patterns[actionLower]) {
-      targetBtn = buttons.find(b => patterns[actionLower].test(getBtnLabel(b)));
-    }
-
-    // 5. Fall back to regex on raw action text
-    if (!targetBtn) {
-      const escapedAction = searchText.replace(/[-\/\\^$*+?.()|[\]{}]/g, '\\$&');
-      const pattern = new RegExp(escapedAction, 'i');
-      targetBtn = buttons.find(b => pattern.test(getBtnLabel(b)));
-    }
-
-    if (targetBtn) {
-      targetBtn.click();
-      
-      // If this is a numbered option (1., 2., 3.), also click Submit after a short delay
-      const clickedText = getBtnLabel(targetBtn);
-      if (/^\d+\./.test(clickedText)) {
-        setTimeout(() => {
-          const submitBtn = buttons.find(b => /^(제출|submit)/i.test(getBtnLabel(b)));
-          if (submitBtn && submitBtn !== targetBtn) submitBtn.click();
-        }, 150);
+    for (const btn of buttons) {
+      const text = (btn.textContent || '').trim();
+      const label = btn.getAttribute('aria-label') || '';
+      if (pattern.test(text) || pattern.test(label)) {
+        btn.click();
+        return JSON.stringify({ success: true, action, clicked: text || label });
       }
-      
-      return JSON.stringify({ success: true, action, clicked: clickedText });
     }
 
     return JSON.stringify({
       success: false,
-      error: `No button matching '${searchText}' found`,
-      available: buttons.map(b => getBtnLabel(b)).filter(t => t.length > 0 && t.length < 80),
+      error: `No button matching action '${action}' found`,
+      available: buttons.map(b => (b.textContent || '').trim()).filter(t => t.length > 0 && t.length < 40),
     });
   } catch (e) {
     return JSON.stringify({ error: e.message || String(e) });

package/providers/_builtin/extension/codex/scripts/1.0/send_message.js

@@ -12,15 +12,35 @@
     
     // Find ProseMirror editor
     const editor = document.querySelector('.ProseMirror');
-    if (!editor) return JSON.stringify({ error: 'Editor not found' });
+    if (!editor) return JSON.stringify({ error: 'ProseMirror editor not found' });
 
     // Focus the editor
     editor.focus();
 
-    // Use execCommand to safely insert text. This avoids TrustedHTML errors 
-    // and naturally triggers ProseMirror's state updates and Keyboard/Input events.
-    document.execCommand('selectAll', false, null);
-    document.execCommand('insertText', false, message);
+    // Clear existing content
+    const existingP = editor.querySelector('p');
+    if (existingP) {
+      existingP.textContent = message;
+      // Dispatch input event for ProseMirror to detect the change
+      editor.dispatchEvent(new InputEvent('input', {
+        bubbles: true,
+        cancelable: true,
+        inputType: 'insertText',
+        data: message,
+      }));
+    } else {
+      // Fallback: create new paragraph
+      const p = document.createElement('p');
+      p.textContent = message;
+      editor.innerHTML = '';
+      editor.appendChild(p);
+      editor.dispatchEvent(new InputEvent('input', {
+        bubbles: true,
+        cancelable: true,
+        inputType: 'insertText',
+        data: message,
+      }));
+    }
 
     // Wait a tick then submit via Enter key
     setTimeout(() => {

package/src/commands/stream-commands.ts

@@ -229,7 +229,15 @@ export async function handleExtensionScript(h: CommandHelpers, args: any, script
     }
 
     const scriptFn = provider.scripts[actualScriptName as keyof typeof provider.scripts] as Function;
-    const scriptCode = scriptFn(args);
+    // Normalize args: script placeholders use UPPERCASE (${MODE}, ${MODEL}, ${MESSAGE})
+    // but WebSocket args typically use lowercase. Add uppercase versions of common keys.
+    const normalizedArgs = { ...args };
+    for (const key of ['mode', 'model', 'message', 'action', 'button', 'text', 'sessionId']) {
+        if (key in normalizedArgs && !(key.toUpperCase() in normalizedArgs)) {
+            normalizedArgs[key.toUpperCase()] = normalizedArgs[key];
+        }
+    }
+    const scriptCode = scriptFn(normalizedArgs);
     if (!scriptCode) return { success: false, error: `Script '${actualScriptName}' returned null` };
 
     const cdpKey = provider.category === 'ide' ? (h.currentIdeType || agentType) : (h.currentIdeType || ideType);
@@ -249,10 +257,35 @@ export async function handleExtensionScript(h: CommandHelpers, args: any, script
                     break;
                 }
             }
-            if (!targetSessionId) {
-                return { success: false, error: `No active session found for ${agentType}` };
+
+            // IDE-level scripts (model/mode) — try session frame first, fallback to main page
+            const IDE_LEVEL_SCRIPTS = ['listModes', 'setMode', 'listModels', 'setModel'];
+            if (IDE_LEVEL_SCRIPTS.includes(scriptName)) {
+                // Try session frame first (some extensions embed mode selector in their webview)
+                if (targetSessionId) {
+                    try {
+                        result = await cdp.evaluateInSessionFrame(targetSessionId, scriptCode);
+                        // Check if result indicates "not found" — fallback to main page
+                        const parsed = typeof result === 'string' ? JSON.parse(result) : result;
+                        const notFound = parsed?.error?.includes('not found') || parsed?.error?.includes('no root');
+                        if (notFound) {
+                            LOG.info('Command', `[ExtScript] ${scriptName} not found in session frame → trying IDE main page`);
+                            result = await cdp.evaluate(scriptCode, 30000);
+                        }
+                    } catch {
+                        LOG.info('Command', `[ExtScript] ${scriptName} session frame failed → trying IDE main page`);
+                        result = await cdp.evaluate(scriptCode, 30000);
+                    }
+                } else {
+                    LOG.info('Command', `[ExtScript] ${scriptName} no session → trying IDE main page`);
+                    result = await cdp.evaluate(scriptCode, 30000);
+                }
+            } else {
+                if (!targetSessionId) {
+                    return { success: false, error: `No active session found for ${agentType}` };
+                }
+                result = await cdp.evaluateInSessionFrame(targetSessionId, scriptCode);
             }
-            result = await cdp.evaluateInSessionFrame(targetSessionId, scriptCode);
         } else if (hasWebviewScript && cdp.evaluateInWebviewFrame) {
             const matchText = provider.webviewMatchText;
             const matchFn = matchText ? (body: string) => body.includes(matchText) : undefined;

package/src/daemon/dev-server.ts

@@ -339,15 +339,29 @@ export class DevServer {
       }
       this.log(`Exec script length: ${scriptCode.length}, first 50 chars: ${scriptCode.slice(0, 50)}...`);
 
-      // Execute webview script via evaluateInWebviewFrame
-      const isWebviewScript = provider.category === 'extension' || scriptName.toLowerCase().includes('webview');
+      // Execute based on provider category
+      const isWebviewScript = scriptName.toLowerCase().includes('webview');
       let raw: any;
-      if (isWebviewScript) {
+      if (provider.category === 'extension' && !isWebviewScript) {
+        // Extension scripts: prefer session frame (agent webview) — matching agent-stream poller behavior
+        const sessions = cdp.getAgentSessions();
+        let sessionId: string | null = null;
+        for (const [sid, target] of sessions) {
+          if (target.agentType === type) { sessionId = sid; break; }
+        }
+        if (sessionId) {
+          raw = await cdp.evaluateInSessionFrame(sessionId, scriptCode);
+        } else if (cdp.evaluateInWebviewFrame) {
+          // Fallback: try evaluateInWebviewFrame
+          const matchText = provider.webviewMatchText;
+          const matchFn = matchText ? (body: string) => body.includes(matchText) : undefined;
+          raw = await cdp.evaluateInWebviewFrame(scriptCode, matchFn);
+        } else {
+          raw = await cdp.evaluate(scriptCode, 30000);
+        }
+      } else if (isWebviewScript && cdp.evaluateInWebviewFrame) {
         const matchText = provider.webviewMatchText;
         const matchFn = matchText ? (body: string) => body.includes(matchText) : undefined;
-        if (!cdp.evaluateInWebviewFrame) {
-            throw new Error(`CDP manager does not support evaluateInWebviewFrame`);
-        }
         raw = await cdp.evaluateInWebviewFrame(scriptCode, matchFn);
       } else {
         raw = await cdp.evaluate(scriptCode, 30000);

package/src/status/reporter.ts

@@ -221,7 +221,7 @@ export class DaemonStatusReporter {
             LOG.debug('P2P', `sent (${JSON.stringify(payload).length} bytes)`);
         }
 
- // ═══ Server transmit (minimal routing meta only — sanitizeForRelay removes everything else) ═══
+ // ═══ Server transmit (minimal routing meta only) ═══
         if (opts?.p2pOnly) return;
         const wsPayload = {
             daemonMode: true,