Delegate quantitative research to AutoQuant

Treat AutoQuant as a specialist coworker, not a function call. OpenAlice owns the conversation and routing; the AutoQuant Agent owns Project selection, research method, dependencies, experiments, Git history, and evidence.

Bound the assignment

Before dispatch, preserve the caller-owned parts of the question:

• the decision the research should support;
• assets or universe, direction, horizon, and cadence when the caller knows them;
• available data and its provenance;
• benchmark, costs, constraints, risk limits, and requested deliverable;
• material unknowns that must be clarified rather than invented.

Do not translate the request into AutoQuant CLI steps, choose a Project or Study on its behalf, promise a positive result, or grant trading authority. A negative result, unsupported route, or request for clarification can be a correct handoff.

Recruit the default desk

Use the alice collaboration surface:

alice conversation create --harness autoquant --await --prompt '
Research question: <question>
Decision this supports: <decision>
Caller-owned scope and constraints: <assets, direction, horizon, cadence, benchmark, costs, limits>
Available evidence or data: <paths and provenance, or say what is missing>
Expected handoff: answer in plain language; name the Project and every applicable immutable Run, Experiment, Report, or Dossier id; give the absolute Project root and end with `Primary deliverable directory: <absolute path>` pointing to the Dossier bundle, Report bundle, or Project directory that contains the named evidence; distinguish validation, visible-test, and external-holdout evidence; state assumptions, unsupported claims, and that trading authority is none.
Ask before proceeding if a missing caller-owned fact would materially change the study. Do not manufacture a Report or Dossier solely for transport.
'