Convert book and video methodologies into executable AI skills
You may have finished a book or podcast but found yourself unable to recall the content when you needed it. This tool converts that knowledge into 'skills' that an AI agent can call directly in actual work. It does not simply summarize content; it extracts verified methodologies and restructures them into an executable form.
Stability improvements in v2.5.0 and latest package verification
The recent v2.5.0 version improves stability by treating the 'Capability Bundle' as the single source of truth. Capability cards and metadata are now generated first during the extraction phase, followed by the compilation of installable outputs. Existing users must check BUILD_INFO.json to update to the latest package, and note that this refresh is not included in GitHub's automatic archives.
The 7-step processing of the RIA-TV++ pipeline
This tool uses the RIA-TV++ pipeline to convert source text into reusable capability bundles. It first applies Mortimer Adler's analysis method to structure the entire content, then five specialized extractors run in parallel to pull out frameworks, principles, and cases. The content then passes through triple verification and a promotion gate to confirm executability, with Zettelkasten-style links coding dependencies between capabilities. Finally, only content that passes a stress test including temptation questions is distributed in single or pack form through a deterministic compilation process.
Knowledge structuring for agents, not humans
Traditional reading notes or summaries are compressed information for human consumption, not structures for AI execution. This tool takes an execution-oriented approach that clearly defines 'when to use what.' Methodologies extracted from various sources such as books, subtitled videos, podcasts, and interviews are converted into testable skill packs that can be called independently and combined. This allows AI agents to actively utilize stored knowledge in real scenarios.
Installation environment and DeepSeek Harness integration
Running the scripts requires Python 3.10 or higher and PyYAML, and you can perform a self-diagnosis with the python3 scripts/cangjie.py doctor command. If you are using DeepSeek Harness, download the adapter layer from the release package and install it as a local tarball. After installation, simply specify the file path and enter a natural language command requesting skill extraction when starting a new task. For video content, it is recommended to first extract subtitles or audio transcripts using the video-downloader skill before passing them to this tool.
Target audience and verification points
This tool is suitable for developers and knowledge managers who want to port systematic methodologies from books or videos to AI agents. It is particularly useful for converting verified business insights, such as Buffett's shareholder letters or Netflix's culture principles, into digital assets. However, not all content is converted into skills due to strict filtering, so the density of extractable methodologies in the source is important. It is released under the MIT license, and external source contributions are included with the author's permission.