Edge Hint Extractor

This extracts structured edge hints from market observations, anomalies, and news reactions, outputting a canonical hints.yaml for downstream concept synthesis. It's the first stage in a four-part workflow: observe, abstract, design, pipeline. You run it rule-based by default, but can augment with LLM-generated ideas either via subprocess or by passing a pre-written YAML file, which is handy if you're using Claude Code and want Claude to draft the hints itself. The file-based approach is a smart workaround for agentic workflows where you want the LLM in the loop without shelling out. Straightforward Python script with clear separation between deterministic extraction and optional creative augmentation.