Deep Research

You can synthesize 30 to 50 sources in a single 1M-token context session using the Deep Research skill. It applies a CRAAP-lite credibility rubric to classify sources by type and assign per-finding confidence levels for evidence-based reports.

Source credibility classification works by applying a CRAAP-lite rubric to tag sources as primary, secondary, or tertiary. This process weights findings by confidence levels to ensure traceability and accurate evidence synthesis.

The best way to generate a research digest with contradictions and falsifiable claims is to run a deep topic investigation. The process outputs a structured report featuring an executive summary, data points, and explicitly falsifiable claims.

You can use this workflow without prior context, but it reads memory/MEMORY.md to incorporate existing context if available. Provide a topic and set the depth parameter to deep to start a comprehensive on-demand investigation.

The structured output of a multi-source research synthesis includes an executive summary, tagged data points, identified contradictions, falsifiable claims sections, and recommended actions based on the evaluated evidence.

A limitation when processing large bodies of evidence is the session capacity constraint of 30 to 50 sources within a 1M-token context. Exceeding this volume may impact the accuracy of per-finding confidence and source traceability.