vtt-to-text

VTT to Text Converter

Convert pasted WebVTT closed caption content into clean, readable plain text files.

Workflow

Step 1: Identify VTT Content

The user will paste VTT content directly into the conversation. Recognizable by:

• A WEBVTT header line
• Numbered sequence blocks (bare integers on their own line)
• Timestamp lines with --> arrows (e.g., 00:00:03.885 --> 00:00:05.395)
• Caption text lines between timestamps

Step 2: Ask for Filename Details

Before processing, ask the user for three things using AskUserQuestion:

1. Tier — e.g., "Tier 1", "Tier 2", "Tier 3"
2. Sprint number — e.g., "Sprint 1", "Sprint 2", "Sprint 14"
3. Title — the class or lecture name, e.g., "Introduction", "Growth Strategy", "Financial Planning"

The final filename pattern is: {Tier} - {Sprint} - {Title}.txt Example: Tier 3 - Sprint 1 - Introduction.txt

If the user has already provided any of these details in their message, skip those questions and use what they gave.

Step 3: Parse and Clean

Save the raw VTT content to a temp file and run the bundled parser:

python3 ${CLAUDE_PLUGIN_ROOT}/skills/vtt-to-text/scripts/parse_vtt.py < /tmp/raw_vtt.txt > /tmp/cleaned.txt

The parser:

• Strips the WEBVTT header
• Removes sequence numbers and timestamp lines
• Strips inline VTT tags
• Joins caption fragments into a continuous text stream
• Breaks into readable paragraphs (roughly every 5 sentences)

Step 4: Save and Deliver

Read the cleaned output, save it to the user's workspace folder with the filename from Step 2, and provide a computer:// link.

Keep the response brief — confirm the file was created and link it. No need to show the full text content.