Trade Hypothesis Ideator takes a structured JSON bundle of trade logs, journal snippets, market data and observations, and turns it into hypothesis cards. Each card is critiqued, ranked and given a verdict of pursue, revise or discard, with an experiment design and kill criteria.
Cards marked pursue can be exported as strategy YAML and a metadata JSON compatible with an edge-finder candidate format. It is a pure calculation and prompting skill that needs no API key and does not give investment advice.
What You Get
- Ranked hypothesis cards in JSON and Markdown with a pursue, revise or discard verdict
- An experiment design and kill criteria on every card
- Optional strategy YAML and metadata JSON for each pursued hypothesis
- A taxonomy and an evidence quality guide to judge sample size and strength
Common Use Cases
- Turning a month of trade journal notes into testable ideas
- Checking a hunch about market behavior against logged evidence before backtesting
- Preparing candidate strategies for a research pipeline with provenance metadata
What Makes It Unique
Every hypothesis must be falsifiable, so each card carries kill criteria that say when to drop the idea. Prompts for generation and critique are separate, and a ranking pass with guardrails runs after the critique, which keeps weak ideas from reaching the export step.
Under The Hood
A Python script run from the shell does the two mechanical passes: it normalizes the input bundle and summarizes the evidence, then ranks and formats the cards. The hypothesis writing and critique steps are done by the agent from prompt files in the skill folder. Output files go into a reports folder, and strategy export writes extra YAML and JSON files. The skill needs PyYAML, makes no network calls and uses no API key.