mj-audio-topics

What problem does it solve?

Manually transcribing hours of Chinese investment audio, extracting topics, and organizing them into a searchable knowledge base is extremely tedious and error-prone. This Skill automates the entire pipeline from local audio files to structured, scored topic documents in Lark Wiki.

Core Features & Use Cases

• Two-Phase Workflow: Phase 1 transcribes audio, extracts topics with a 5-dimension scoring model, and appends summaries to topic documents. Phase 2 performs deep web search and comprehensive analysis on high-score topics selected by the user.
• Multi-Tagging & State Management: A single topic can be appended to multiple relevant topic documents (for example, Japan military spending to both Japan Market and Defense and Aerospace). Atomic state files track processed files, topic registries, and support resumable batch processing.
• Use Case: An investor like MJ who follows multiple weekly audio series like 时寒冰微课 can automatically build an evolving topic evolution axis knowledge base without manual transcription or document management.

Quick Start

Place your Chinese investment audio files in ~/Downloads/Claude-Audio-Input/<series>/ and invoke the /mj-audio-topics command to automatically transcribe, score topics, and build your structured Lark Wiki knowledge base.