autoskill

Automating repetitive research workflows involves analyzing local screen activity to detect and cluster repeated patterns. The system observes your tasks and automatically proposes new AI skills or composition recipes to streamline future work.

Yes, screen recording data is processed locally when detecting repeated research patterns. All analysis, clustering, and redaction of sensitive information occur directly on your machine to ensure privacy-first workflow detection.

Yes, a local screenpipe daemon is required to synthesize AI skills from your daily tasks. The system also requires sentence-transformers for embedding generation and a configured LLM backend for final skill synthesis.

Yes, you can generate new AI skills for scientific and technical research using workflow detection. The system specifically targets these domains by matching observed screen patterns against existing library capabilities to cover toolkit gaps.

To run autoskill and draft new skill proposals, invoke the autoskill run command with your desired time window. This analyzes your recent local screen activity and generates new skill proposals based on detected patterns.