research-loop

Automating unattended research loops for quantitative trading involves running reversible experiments and decision-making cycles without human interaction. This Skill executes iterative phases including brainstorming, spec, plan, execution, verdict, and merge autonomously.

An autonomous research loop is a hands-off cycle for feature, signal, and validation research. It applies pre-defined rules and triggers to execute computational experiments, run subagent-driven execution, and produce verdicts automatically.

Yes, you can run reversible computational experiments without human interaction. The automation focuses specifically on reversible actions and decisions, applying strict rules to prevent irreversible actions and high-stakes judgments.

To start an autonomous research cycle for feature and signal research, you invoke the research-loop command. This triggers the automated workflow, handling brainstorming through the final merge cycle without further input.

The limitation of autonomous decision-making in quantitative trading research is that it restricts irreversible actions and high-stakes judgments. It is designed strictly for reversible experiments, meaning unsafe or permanent trading decisions require manual oversight.

Yes, the research automation supports subagent-driven execution for validation research. It integrates this execution method directly into the unattended loop to carry out computational experiments and deliver automated verdicts.