Video is a rich source of procedural knowledge and a poor format for an agent that needs to apply the same procedure tomorrow. SkillTube attempted to compile that knowledge into a reusable SKILL.md rather than repeatedly asking a model to rediscover it.
A Streamlit application that processes a YouTube source into structured, reusable instructions for an AI-agent skill.
The aim. Convert long-form video knowledge into a structured SKILL.md that an AI agent can reuse.
01The problem behind the project
Video contains useful procedures but is difficult for coding agents to retrieve and apply repeatedly. A compiled skill can make the knowledge operational.
The distinction is between summarization and operationalization. A useful skill must preserve the steps, conditions, tools, and boundaries an agent needs to act, not merely produce a shorter description of the video.
Agent users, educators, and developers may benefit. Video creators and rights holders are affected by transcription, transformation, and redistribution choices.
02How it took shape
A working Streamlit interface, video-processing pipeline, model-assisted synthesis, and SKILL.md output.
The working Streamlit application processed YouTube material, used a model to synthesize structured instructions, and emitted a reusable skill artifact. The product functioned, but it did not find a convincing acquisition path as a standalone service.
Josiah identified the video-to-agent-knowledge opportunity, directed development, evaluated the product, and later incorporated the workflow lesson into AgentWorkbench.
The application worked technically but was not publicly launched and did not establish a viable user-acquisition path.
03What the project means now
The capability survived the product. Its most useful ideas now belong inside AgentWorkbench, where video processing is one discoverable tool among others. SkillTube demonstrates how a failed go-to-market direction can still produce durable infrastructure.
Output quality depends on the video and model, generated instructions require review, and source rights must be respected.
A useful tool can still be a weak standalone product; the same capability may be more valuable as infrastructure inside a broader workbench.
Preserve the working method and fold the best parts into AgentWorkbench rather than reviving the original acquisition problem.