Edge Signal Aggregator reads the local JSON and YAML outputs of several upstream skills: edge-candidate-agent, edge-concept-synthesizer, theme-detector, sector-analyst, institutional-flow-tracker and edge-hint-extractor. It applies configurable weights, merges overlapping signals and ranks the results by composite score.
Each ranked idea lists the skills that contributed to it, and conflicting signals are flagged for manual review. The output is a dashboard of research signals and gives no investment advice.
What You Get
- A timestamped JSON report with ranked signals, contributing skills, contradictions, a deduplication log and invalid signals
- A Markdown dashboard with the top ideas, their tickers, direction and horizon, and the contradictions to review
- A scale assumptions report for scores whose scale was inferred
- Default weights in a YAML file you can copy and change
Common Use Cases
- Seeing which ideas several analysis skills agree on
- Finding conflicts between a sector view and an institutional-flow view
- Choosing which edge ideas to research further by composite score
What Makes It Unique
It treats scoring strictly: NaN, infinite, negative and out-of-range values are rejected into a report and never converted to high confidence, and the newest observation wins when copies of the same evidence differ. Contradictions are listed in the report, and a documented rule limits agreement credit to distinct events from distinct skills.
Under The Hood
The agent runs one local Python script that reads report files from the reports folder and a weights YAML, and writes JSON and Markdown reports back to the output directory. It needs Python and PyYAML, no API keys, and makes no web calls. The agent runs shell commands and writes files. A reference document on the weighting framework loads when needed.