Subagent Driven Analysis

When you have a frozen, pre-registered analysis plan with independent steps, this skill runs them one by one using fresh subagents per step, each with isolated context to prevent improvisation. After each step completes, it dispatches two reviewers in sequence: first a protocol compliance check to ensure nothing beyond the pre-registration was run, then a statistical rigor review for correctness and reproducibility. The continuous execution loop only pauses for blockers or genuine ambiguities. This is rigorous to the point of being almost bureaucratic, but that's the design. If your analysis isn't pre-registered yet or your harness doesn't support subagents, you'll need different skills first.

Install to Claude Code

npx -y skills add k-dense-ai/science-superpowers --skill subagent-driven-analysis --agent claude-code