Edge Concept Synthesizer sits between detection and strategy design in the edge research workflow. It reads ticket YAML files from a detector run and an optional hints file, groups tickets that share a mechanism, and writes edge_concepts.yaml.
Each concept carries support statistics, an abstract thesis, invalidation signals and an export-readiness flag. Overlapping concepts for the same hypothesis are merged. The output is research structure and gives no investment advice.
What You Get
- An edge_concepts.yaml file with concept clusters
- Support statistics for each concept, from the tickets behind it
- An abstract thesis and a list of invalidation signals per concept
- An export-readiness flag that tells the next stage which concepts can move on
Common Use Cases
- Reducing a large pile of raw detector tickets to a short list of mechanisms
- Reviewing concepts before any strategy is drafted, to limit overfitting to single tickets
- Merging near-duplicate ideas that came from different detector runs
What Makes It Unique
It adds an abstraction layer that other edge skills skip: strategies are drafted from reviewed concepts, and a single ticket never maps straight to a strategy. Deduplication uses a containment threshold you can tune or turn off, and a synthetic-ratio cap limits how much promoted hint material can outweigh real tickets.
Under The Hood
The agent runs one local Python script that reads YAML ticket files and an optional hints file from disk and writes edge_concepts.yaml to the path you give. It needs Python and PyYAML. The skill does not call web services, and the agent only runs shell commands and writes the output file. A schema reference loads when needed.