Verified Study Analysis

Dual-agent study analysis with cross-verification.

Overview

Two agents analyze the same data in parallel and independently. A third pass cross-verifies their results for convergence and flags discrepancies before producing a final summary plus annotated R or Python code that you re-run to verify the analysis yourself. Running two agents in parallel surfaces silent coding errors a single agent would miss. A pre-analysis interview detects whether your documentation is complete (codebook, preregistration, analysis template). If everything is in place, it runs a one-shot confirmation (FAST-PATH); if not, it runs a four-area Q&A (INTERVIEW) covering variable interpretation, condition coding, primary analyses, and detail level. Bundles study-summary (output templates and AMA formatting) and methodological-detail-log (audit trail).

When to use

• Getting annotated R or Python code that you re-run to verify the analysis yourself
• Catching silent coding errors a single agent would miss, by running two agents in parallel and flagging where they disagree
• Re-running an analysis from a coauthor for a verified second pass
• Building a reviewer-defensible audit trail of every analytical decision
• Picking up a project mid-stream so the agent reads its own context instead of waiting for a manual brief

Download

Unzip into ~/.claude/skills/ (Claude Code) or ~/.codex/skills/ (Codex), then restart your CLI.

Feedback

Found a bug or have a suggestion? Email me.