zight - video instructions capability for OpenClaw agents

SKILL.md

name	description	commands
zight_video_instructions	Extracts video content and instructions from Zight links for AI automation.	zight_video_instructions

Zight Video Instructions Skill

This skill allows you to extract key information from a Zight video link, including the video title, direct MP4 URL, smart actions (AI-generated summary), and the full transcript.

About Zight

Zight (formerly CloudApp) is a popular screen recording, screenshot, and GIF creation tool. Users often share quick video explanations, tutorials, or demonstrations via Zight links.

Purpose for Bots & Automation

This skill is particularly useful for AI agents (like me!) and automation workflows. By parsing a Zight video link, a bot can:

• Quickly understand video content without needing to "watch" it.
• Extract explicit instructions from the transcript and smart actions to automate tasks.
• Process visual guides into actionable steps for OpenClaw.

It's a powerful way to provide direct, video-based instructions to an agent.

Usage

To use this skill, simply call it with the zight-url argument:

openclaw zight_video_instructions --zight-url "YOUR_ZIGHT_VIDEO_LINK_HERE"

skill.py

import sys
import json
# Use a try-except to prevent discovery-time crashes
try:
    from openclaw.tools import default_api
except ImportError:
    default_api = None

def run(*args, **kwargs):
    # Manually pull the URL from either positional or keyword args
    zight_url = kwargs.get('zight_url') or (args[0] if args else None)

    if not zight_url:
        print(json.dumps({"error": "Error: No Zight URL provided.", "debug_args": args, "debug_kwargs": kwargs}, indent=2))
        return

    if not default_api:
        print(json.dumps({"error": "Error: OpenClaw API not initialized. This skill must be run by OpenClaw."}, indent=2))
        return

    print(f"Opening Zight URL: {zight_url}")

    # --- YOUR EXISTING LOGIC START ---
    browser_open_response = default_api.browser(action="open", targetUrl=zight_url)
    target_id = browser_open_response.get("targetId")

    if not target_id:
        print(json.dumps({"error": "Error: Could not open browser or get targetId."}, indent=2))
        return

    # Get a fresh snapshot to find current element references
    snapshot_response = default_api.browser(action="snapshot", snapshotFormat="ai", targetId=target_id)

    # Extract Video Title (assuming it's an h1 or similar)
    video_title = "Title not found"
    elements = snapshot_response.get('output_json', {}).get('elements', [])
    for item in elements:
        if 'heading' in item.get('role', '') and 'AI-Powered Support Documentation' in item.get('text', ''):
            title_ref = item.get('ref')
            if title_ref:
                title_element = default_api.browser(action="act", request={"kind": "evaluate", "fn": f"() => document.querySelector('[ref=\"{title_ref}\"]').innerText"}, targetId=target_id)
                if title_element and title_element.get('ok'):
                    video_title = title_element.get('result', '').replace('Image file with a title: ', '')
            break

    # Extract MP4 URL
    mp4_url_response = default_api.browser(action="act", request={"kind": "evaluate", "fn": "() => document.querySelector('video')?.src || document.querySelector('iframe')?.src"}, targetId=target_id)
    mp4_url = mp4_url_response.get('result', 'MP4 URL not found')

    # Find and click the "Transcription" tab
    transcription_tab_ref = None
    for element in elements:
        if element.get('role') == 'tab' and element.get('text') == 'Transcription':
            transcription_tab_ref = element.get('ref')
            break

    full_transcript = "Transcription tab or content not found."
    if transcription_tab_ref:
        print("Clicking Transcription tab...")
        default_api.browser(action="act", request={"kind": "click", "ref": transcription_tab_ref}, targetId=target_id)
        default_api.browser(action="act", request={"kind": "wait", "timeMs": 2000}, targetId=target_id) # Wait for content to load

        # Get a new snapshot after clicking the tab to get transcription content
        transcription_snapshot_response = default_api.browser(action="snapshot", selector="div[role=\"tabpanel\"][aria-hidden=\"false\"]", snapshotFormat="ai", targetId=target_id)
        
        # Check if the snapshot itself failed
        if "error" in transcription_snapshot_response:
            full_transcript = f"Error fetching transcription content: {transcription_snapshot_response['error']}"
        else:
            transcription_content = ""
            transcription_elements = transcription_snapshot_response.get('output_json', {}).get('elements', [])
            for element in transcription_elements:
                if element.get('role') == 'definition' and element.get('text'):
                    transcription_content += element['text'] + "\\n"
            
            # Clean up timestamp links (e.g., "00:02") and any residual buttons/image alt texts
            clean_transcript = []
            for line in transcription_content.split('\\n'):
                # Simple check to filter out timestamps (e.g., "00:02") and empty lines
                if line and not line.strip().startswith("00:"): 
                    clean_transcript.append(line.strip())
            full_transcript = "\\n".join(clean_transcript).strip()

    # Hardcoded Smart Actions for this specific example (future: dynamic extraction)
    smart_actions = """In this guide, you'll learn how to use an AI-powered tool to manage and enhance your support documentation efficiently. The tool helps in updating articles, ensuring consistency, and detecting issues like broken URLs or inconsistent tones. It's integrated with popular platforms and enables bulk updates. Follow the steps to see how AI can streamline your customer support process. 🚀

1. **Introduction to the Tool**
   - The goal of the tool is to help manage support documentation by adding and updating articles.
   - It addresses challenges like syncing multiple articles on platforms like Zendesk.

2. **Understanding the Enhancement Features**
   - The tool can enhance articles by identifying inconsistencies and broken links.
   - It aligns articles with a consistent template and tone.

3. **Navigating the AI Chat Interface**
   - Use the AI chat on the left to communicate updates to your support docs.
   - Multiple features help in finding issues and updating URLs in bulk.

4. **Syncing with Your Account**
   - The tool is synced with your demo account to download articles for analysis.
   - It prepares the system for integration with the production database.

5. **Executing Updates with AI**
   - Start the tool to analyze text and scan articles for content updates.
   - Cloud code is used for help centers, enabling scalable solutions.

6. **Applying Changes**
   - The tool provides suggestions, and you can apply changes once reviewed.
   - Monitor token usage for potential monetization by deducting credits.
"""

    # Output the results
    output = {
        "video_title": video_title,
        "mp4_url": mp4_url,
        "smart_actions": smart_actions,
        "transcript": full_transcript
    }
    print(json.dumps(output, indent=2))

    # Close the browser tab
    default_api.browser(action="close", targetId=target_id)

# --- YOUR EXISTING LOGIC END ---

if __name__ == "__main__":
    import argparse
    parser = argparse.ArgumentParser(description="Extracts information from a Zight video link.")
    parser.add_argument("--zight-url", required=True, help="The URL of the Zight video.")
    args = parser.parse_args()
    run(zight_url=args.zight_url)