Edge Strategy Reviewer checks the strategy_drafts YAML files produced by edge-strategy-designer before they go to export. A Python script scores each draft on eight weighted criteria: edge plausibility, overfitting risk, sample adequacy, regime dependency, exit calibration, risk concentration, execution realism and invalidation quality.
The result is a verdict with a confidence score and, for REVISE, revision instructions. It reviews research drafts for internal quality and gives no investment advice.
What You Get
- A review.yaml or review.json with a verdict, confidence score, findings and revision instructions for every draft
- A summary count of PASS, REVISE and REJECT and of export-eligible drafts
- An optional Markdown summary
- A bundled bias checklist you can swap for your own YAML file
Common Use Cases
- Gating drafts before they are exported to a backtest pipeline
- Reading the revision instructions for a weak draft
- Checking a single hand-written draft for overfitting and sample-size problems
What Makes It Unique
The review is deterministic, with fixed weights and thresholds, so the same draft gets the same verdict. Overfitting is scored by filter count tier and by a penalty for precise thresholds, and the bias checklist option adds a research-bias gate that can remove export eligibility.
Under The Hood
The agent runs one local Python script that reads draft YAML files and writes review output files to the directory you choose. It needs Python and PyYAML, uses no API keys, and the skill describes no web calls. The agent runs shell commands and writes files. Reference documents with the scoring rubric and the overfitting checklist load when needed.