@adhdev/daemon-core 0.5.42 → 0.5.44

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@adhdev/daemon-core",
-  "version": "0.5.42",
+  "version": "0.5.44",
   "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*this\\s*command", "flags": "i" },
+      { "source": "Run\\s+\\w+\\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

@@ -1,88 +1,58 @@
 /**
- * Codex Extension — Deep DOM Explorer
- * Explores the chat area, message list, input field, model selector etc.
+ * Debug script — enumerate all buttons in all frames
  */
 (() => {
   try {
-    const root = document.getElementById('root');
-    if (!root) return JSON.stringify({ error: 'no root' });
-
-    // 1. Find chat input (textarea, contenteditable, etc.)
-    const textareas = document.querySelectorAll('textarea');
-    const contentEditables = document.querySelectorAll('[contenteditable="true"]');
-    const chatInputCandidates = document.querySelectorAll('[data-placeholder], [placeholder*="ask"], [placeholder*="message"], [placeholder*="type"]');
-
-    // 2. Find message-like containers
-    const messageContainers = [];
-    // Look for common message list patterns
-    const candidates = [
-      ...document.querySelectorAll('[class*="message"]'),
-      ...document.querySelectorAll('[class*="chat"]'),
-      ...document.querySelectorAll('[class*="conversation"]'),
-      ...document.querySelectorAll('[class*="thread"]'),
-      ...document.querySelectorAll('[role="log"]'),
-      ...document.querySelectorAll('[role="list"]'),
-    ];
-    for (const el of candidates) {
-      messageContainers.push({
-        tag: el.tagName?.toLowerCase(),
-        class: (el.className && typeof el.className === 'string') ? el.className.substring(0, 200) : null,
-        role: el.getAttribute('role'),
-        childCount: el.children?.length || 0,
-        text: (el.textContent || '').trim().substring(0, 200),
-      });
+    const results = {};
+    
+    // Frame 1: document (child frame / execution context)
+    const allBtns1 = Array.from(document.querySelectorAll('button'));
+    const visibleBtns1 = allBtns1.filter(b => b.offsetWidth > 0);
+    const menuBtns1 = visibleBtns1.filter(b => b.getAttribute('aria-haspopup') === 'menu');
+    results.childFrame = {
+      hasRoot: !!document.getElementById('root'),
+      totalBtns: allBtns1.length,
+      visibleBtns: visibleBtns1.length,
+      menuBtns: menuBtns1.length,
+      menuBtnTexts: menuBtns1.map(b => (b.textContent||'').trim().slice(0,40)),
+      sampleBtnTexts: visibleBtns1.slice(0,10).map(b => ({
+        t: (b.textContent||'').trim().slice(0,40),
+        hp: b.getAttribute('aria-haspopup'),
+        w: b.offsetWidth
+      }))
+    };
+    
+    // Frame 2: inner iframe (if available)
+    const iframes = document.querySelectorAll('iframe');
+    results.iframeCount = iframes.length;
+    for (let i = 0; i < Math.min(iframes.length, 3); i++) {
+      try {
+        const innerDoc = iframes[i].contentDocument || iframes[i].contentWindow?.document;
+        if (innerDoc) {
+          const allBtns2 = Array.from(innerDoc.querySelectorAll('button'));
+          const visibleBtns2 = allBtns2.filter(b => b.offsetWidth > 0);
+          const menuBtns2 = visibleBtns2.filter(b => b.getAttribute('aria-haspopup') === 'menu');
+          results['innerFrame_' + i] = {
+            hasRoot: !!innerDoc.getElementById('root'),
+            url: iframes[i].src?.slice(0, 100),
+            totalBtns: allBtns2.length,
+            visibleBtns: visibleBtns2.length,
+            menuBtns: menuBtns2.length,
+            menuBtnTexts: menuBtns2.map(b => (b.textContent||'').trim().slice(0,40)),
+            sampleBtnTexts: visibleBtns2.slice(0,10).map(b => ({
+              t: (b.textContent||'').trim().slice(0,40),
+              hp: b.getAttribute('aria-haspopup'),
+              w: b.offsetWidth
+            }))
+          };
+        }
+      } catch(e) {
+        results['innerFrame_' + i] = { error: e.message };
+      }
     }
-
-    // 3. Find model/mode selectors
-    const modelCandidates = document.querySelectorAll('[class*="model"], [class*="gpt"], [aria-label*="model" i], [aria-label*="모델" i]');
-
-    // 4. Enumerate all role="button" items (conversation list items)
-    const roleButtons = document.querySelectorAll('[role="button"]');
-
-    // 5. Look for "GPT" text or model name in the UI
-    const allText = root.textContent || '';
-    const modelMatch = allText.match(/(GPT-[\d.]+|gpt-[\w.-]+|o\d+-[\w]+|claude-[\w.-]+)/i);
-
-    // 6. Check if we're on a task list or chat view
-    const hasConversationList = document.querySelectorAll('[role="button"][class*="rounded-lg"]').length > 0;
-    const headerText = document.querySelector('[style*="view-transition-name: header-title"]')?.textContent?.trim() || '';
-
-    return JSON.stringify({
-      headerText,
-      isTaskList: headerText === '작업' || headerText === 'Tasks',
-      modelFound: modelMatch ? modelMatch[0] : null,
-      textareaCount: textareas.length,
-      textareas: Array.from(textareas).map(t => ({
-        class: t.className?.substring(0, 200),
-        placeholder: t.placeholder?.substring(0, 200),
-        rows: t.rows,
-        value: (t.value || '').substring(0, 100),
-        rect: (() => { const r = t.getBoundingClientRect(); return { w: Math.round(r.width), h: Math.round(r.height) }; })(),
-      })),
-      contentEditableCount: contentEditables.length,
-      contentEditables: Array.from(contentEditables).slice(0, 5).map(el => ({
-        tag: el.tagName?.toLowerCase(),
-        class: (el.className && typeof el.className === 'string') ? el.className.substring(0, 200) : null,
-        text: (el.textContent || '').trim().substring(0, 100),
-        placeholder: el.getAttribute('data-placeholder')?.substring(0, 100),
-      })),
-      chatInputCandidates: Array.from(chatInputCandidates).slice(0, 5).map(el => ({
-        tag: el.tagName?.toLowerCase(),
-        class: (el.className && typeof el.className === 'string') ? el.className.substring(0, 200) : null,
-        placeholder: el.getAttribute('placeholder') || el.getAttribute('data-placeholder'),
-      })),
-      messageContainers: messageContainers.slice(0, 15),
-      modelCandidates: Array.from(modelCandidates).slice(0, 5).map(el => ({
-        tag: el.tagName?.toLowerCase(),
-        class: (el.className && typeof el.className === 'string') ? el.className.substring(0, 200) : null,
-        text: (el.textContent || '').trim().substring(0, 100),
-      })),
-      conversationButtons: Array.from(roleButtons).slice(0, 5).map(b => ({
-        class: (b.className && typeof b.className === 'string') ? b.className.substring(0, 200) : null,
-        text: (b.textContent || '').trim().substring(0, 150),
-      })),
-    });
-  } catch (e) {
-    return JSON.stringify({ error: e.message || String(e) });
+    
+    return JSON.stringify(results);
+  } catch(e) {
+    return JSON.stringify({error: e.message});
   }
 })()

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

@@ -0,0 +1,140 @@
+/**
+ * Codex Extension — list_modes
+ *
+ * Finds the mode / autonomy dropdown next to the model chip in the composer footer,
+ * opens it (Radix), reads options, closes. UI expects `modes` + `current` (see ModelModeBar).
+ *
+ * Uses the same multi-searchRoot strategy as readChat for reliable mode button discovery.
+ */
+(() => {
+  try {
+    function resolveDoc() {
+      let doc = document;
+      let root = doc.getElementById('root');
+      if (!root) {
+        const iframes = doc.querySelectorAll('iframe');
+        for (const iframe of iframes) {
+          try {
+            const innerDoc = iframe.contentDocument || iframe.contentWindow?.document;
+            if (innerDoc?.getElementById('root')) {
+              doc = innerDoc;
+              root = innerDoc.getElementById('root');
+              break;
+            }
+          } catch (e) { /* cross-origin */ }
+        }
+      }
+      return { doc, root };
+    }
+
+    function isModelText(text) {
+      return /^(GPT-|gpt-|o\d|claude-|sonnet|opus)/i.test(text);
+    }
+
+    /**
+     * Find the mode menu button using the same multi-root strategy as readChat.
+     * Searches multiple container scopes and both doc frames.
+     */
+    function findModeButton(doc) {
+      for (const d of [doc, document]) {
+        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) {
+          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();
+            // Skip model buttons — we want the mode button
+            if (!isModelText(text) && text.length > 0 && text.length < 30) {
+              return btn;
+            }
+          }
+        }
+      }
+      return null;
+    }
+
+    function openMenu(btn) {
+      const rect = btn.getBoundingClientRect();
+      const cx = rect.left + rect.width / 2;
+      const cy = rect.top + rect.height / 2;
+      btn.dispatchEvent(
+        new PointerEvent('pointerdown', { bubbles: true, clientX: cx, clientY: cy, pointerId: 1 }),
+      );
+      btn.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, clientX: cx, clientY: cy }));
+      btn.dispatchEvent(
+        new PointerEvent('pointerup', { bubbles: true, clientX: cx, clientY: cy, pointerId: 1 }),
+      );
+      btn.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, clientX: cx, clientY: cy }));
+      btn.dispatchEvent(new MouseEvent('click', { bubbles: true, clientX: cx, clientY: cy }));
+    }
+
+    const { doc, root } = resolveDoc();
+    if (!root) return JSON.stringify({ modes: [], current: '', currentMode: '', error: 'no root' });
+
+    const modeBtn = findModeButton(doc);
+    if (!modeBtn) {
+      return JSON.stringify({
+        modes: [],
+        current: '',
+        currentMode: '',
+        error: 'mode menu button not found',
+      });
+    }
+
+    const menuDoc = modeBtn.ownerDocument || doc;
+    const currentLabel = (modeBtn.textContent || '').trim();
+    openMenu(modeBtn);
+
+    return new Promise((resolve) => {
+      setTimeout(() => {
+        let menu = menuDoc.querySelector('[role="menu"][data-state="open"]');
+        if (!menu) {
+          menu = menuDoc.querySelector('[role="menu"]');
+        }
+
+        const collected = [];
+        if (menu) {
+          const items = menu.querySelectorAll(
+            '[role="menuitem"], [role="menuitemradio"], [role="option"], div[class*="cursor-interaction"]',
+          );
+          for (const item of items) {
+            const text = (item.textContent || '').trim();
+            if (
+              text &&
+              text.length > 0 &&
+              text.length < 80 &&
+              !/^모델|^model\b|^select\b/i.test(text)
+            ) {
+              collected.push(text);
+            }
+          }
+        }
+
+        menuDoc.dispatchEvent(
+          new KeyboardEvent('keydown', {
+            key: 'Escape',
+            code: 'Escape',
+            keyCode: 27,
+            bubbles: true,
+          }),
+        );
+
+        const modes = [...new Set(collected)];
+        const out = {
+          modes: modes.length > 0 ? modes : currentLabel ? [currentLabel] : [],
+          current: currentLabel,
+          currentMode: currentLabel,
+        };
+        resolve(JSON.stringify(out));
+      }, 550);
+    });
+  } catch (e) {
+    return JSON.stringify({ error: e.message || String(e), modes: [], current: '', currentMode: '' });
+  }
+})();

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

@@ -52,6 +52,11 @@
     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']);
@@ -291,24 +296,145 @@
 
     // ─── 3. Status ───
     let status = 'idle';
-    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'))) {
+    // 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'))) {
       status = 'generating';
     }
 
-    // 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';
+    // ─── 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;
     }
+    // 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;
@@ -316,30 +442,53 @@
     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 = '';
-    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;
-    }
+    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;
+            }
+          }
+        }
 
-    // ─── 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)],
-        };
+        // 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;
+            }
+          }
+        }
+
+        if (model) break; // found model, no need to keep searching this scope
       }
+      if (model) break; // found in this frame
     }
 
     // ─── 6. Task info ───

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

@@ -2,39 +2,79 @@
  * Codex Extension — resolve_action
  *
  * Clicks approval/denial buttons in the Codex UI.
- * Actions: "approve", "deny", "cancel"
+ * Actions: "approve", "deny", "cancel", or raw button text
  *
- * Placeholder: ${ACTION}
+ * Placeholders: ${action}, ${button}
  */
 (() => {
   try {
-    const action = ${ACTION};
+    const action = ${action};
+    const buttonText = ${button};
     
-    const buttons = Array.from(document.querySelectorAll('button'))
-      .filter(b => b.offsetWidth > 0);
+    // 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]'));
 
-    // Map action to button text patterns
+    const actionLower = (action || '').toLowerCase();
     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,
     };
 
-    const pattern = patterns[action.toLowerCase()] || patterns.approve;
+    // Determine what to search for: use buttonText if provided, otherwise match action pattern
+    const searchText = buttonText || action || '';
+    let targetBtn = null;
 
-    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 });
+    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);
       }
+      
+      return JSON.stringify({ success: true, action, clicked: clickedText });
     }
 
     return JSON.stringify({
       success: false,
-      error: `No button matching action '${action}' found`,
-      available: buttons.map(b => (b.textContent || '').trim()).filter(t => t.length > 0 && t.length < 40),
+      error: `No button matching '${searchText}' found`,
+      available: buttons.map(b => getBtnLabel(b)).filter(t => t.length > 0 && t.length < 80),
     });
   } catch (e) {
     return JSON.stringify({ error: e.message || String(e) });