transcript-organizer

Provide the transcript file path and answer four questions about type, date, topic, and save location. The Skill reads the entire file, applies the meeting template with discussion points, decisions, and action items, then saves a structured Markdown document.

The Skill detects STT files by repeated misrecognition patterns and corrects them using context-based analysis rather than a fixed dictionary. Corrections are applied automatically and listed transparently in a correction notes table in the final document.

Yes, it detects file encoding with the file command and converts UTF-16 or EUC-KR files to UTF-8 using iconv via the included encoding-detector.sh script. The converted file is written to /tmp before analysis begins.

Files above 110,000 tokens are read in chunks, but the entire content is always analyzed end to end. Output requirements scale with length: 300 lines minimum under 1 hour, 500 lines for 1-2 hours, and 700 lines for 3 hours or more.

You specify a project path during the four-question setup, or the Skill suggests folders found via keyword search in the workspace. If no match exists, the document defaults to the ./00-inbox/ folder with a YYYY-MM-DD_topic.md filename.