Action Item Extractor

WHAT IT DOES

Automatically extracts action items, todos, and tasks from any webpage using pattern matching. The script scans the page text for phrases like "need to", "must", "should", "will", "action item", "TODO", "FIXME", and "next steps", identifying actionable content that might otherwise be missed. Each extracted item includes the full text, the pattern that triggered the extraction, and any associated owner mentioned (e.g., "John needs to..." captures "John" as the owner). Items are displayed in a structured list with checkboxes to mark completion. You can export extracted items as text or JSON, and configure custom patterns to match your specific workflow.

WHERE IT RUNS

Works on all websites (match pattern *://*/*), making it perfect for meeting notes, project documentation, email threads, code review comments, technical specifications, or any web page containing action-oriented text. Each page gets its own independent extraction result set, allowing you to track action items separately for different contexts. The extraction results are saved per URL or custom identifier.

HOW TO USE

Install the script and the action item extractor appears in the panel. Click "Extract Now" to scan the current page for action items. Items appear with checkboxes for tracking completion. Review the extracted items, mark them complete as you work through them. Add custom extraction patterns to match your team's specific language (e.g., "assigned to", "responsible for"). Use the export buttons to copy the list as text or JSON. Toggle "Extract on page load" for automatic scanning. The script saves your progress automatically as you work.

REAL MECHANISM: PATTERN-BASED TEXT EXTRACTION WITH OWNER DETECTION AND DEDUPLICATION

This script uses a multi-stage extraction pipeline that combines regex pattern matching with contextual analysis. The getPageText() function first clones the DOM and removes non-content elements (script, style, iframe, aria-hidden elements) using querySelectorAll and remove(), ensuring only visible page text is scanned. The extractActionItems() function splits the text into sentences using regex /[.!?;]\s+/, then tests each sentence against user-configurable patterns stored as {regex, label} objects. When a match is found, the script attempts to extract an owner using a prioritized list of regex patterns targeting common ownership language (e.g., "by", "from", "assigned to", "@ mentions", "[Name] will", "[Name] needs to"). The extraction results are deduplicated when merging with existing items by comparing the first 50 characters of the fullText property, ensuring the same action item isn't duplicated across multiple extraction runs. Items are stored in a persistent data structure keyed by page ID (URL path or custom identifier) with a storage key format of "action_items_" plus the page ID. The render function displays items in chronological order with status badges indicating the matched pattern, owner tags, and completion checkboxes. All operations are reactive, with automatic persistence through saveExtractedItems() and re-render.

THE PANEL

Drag the panel anywhere on screen by clicking and holding the header. It snaps to screen edges when released. Collapse it to a small bubble by clicking the minimize button, and expand it back by clicking the bubble. Resize the panel from any corner. Press Escape to close the configuration panel. Your position and size preferences are saved automatically per site.

PLEASE NOTE

This script requires the free BotGentz browser extension which provides the BGPanel framework and storage APIs. All data (extracted items, patterns, and page identifiers) is stored locally in your browser using GM_setValue and never sent anywhere. No analytics, no tracking, no external calls. The script only reads the current page content to extract action items and never accesses user data, passwords, or sensitive information. Extraction results are saved per page identifier, so ensure you use a unique identifier for each page or context you track.