Executing Analysis

This is the execution half of the science superpowers workflow. You load your frozen pre-registration, review it critically for feasibility issues, then run every step exactly as written with validation checkpoints throughout. The key discipline is stopping immediately when results look weird or you're tempted to deviate from the plan, because silent tweaks destroy confirmatory status. It includes built-in verification steps and pauses after each major section to report back. The docs emphasize this is the fallback for platforms without subagents, since fresh eyes catch more protocol violations during review. If you're doing any analysis where you want to distinguish confirmatory findings from exploratory ones, the rigor here is the whole point.

Install to Claude Code

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