Convert goals, hypotheses, and experiments into an evidence-based decision loop.
npx skills add https://github.com/davdunc/pai-framework --skill science-davdunc
Please help me install this Agent Skill. Skill: Science Source: https://github.com/davdunc/pai-framework/tree/main/skills/Science Command: npx skills add https://github.com/davdunc/pai-framework --skill science-davdunc
Science helps you replace vague intuition with goal-driven, falsifiable investigation so you can decide what to do next based on evidence rather than confirmation bias.
Use cases include debugging complex systems, validating product or prompt changes via structured evaluation, and conducting research-grade investigations across micro/meso/macro timescales.
Use the Science skill to run QuickDiagnosis on the issue “users still get a 500 error during login” by generating at least three hypotheses, testing the fastest to verify, and updating the plan based on what you observe.
High-intent search queries and answers about installing and using this skill.
Experiment design for debugging requires generating a minimum of three falsifiable hypotheses and defining pre-committed success criteria before testing. This structured investigation replaces vague intuition with an evidence-based decision loop that measures results objectively and plans iterative next actions.
Prompt evaluation using the scientific method applies goal-first success criteria to define measurable thresholds and anti-goals. By generating multiple testable hypotheses, you can measure prompt outputs against pre-committed constraints and analyze the evidence to decide whether to iterate or ship.
A quick diagnosis for complex system errors uses a shortcut method designed for under fifteen minutes. It accelerates debugging by generating at least three hypotheses, testing the fastest one to verify, and updating the action plan based on the observed evidence.
Structured investigation supports research investigations across micro, meso, and macro scales spanning minutes to months. It handles multi-factor problems by enforcing hypothesis plurality, pre-committed success criteria, structured measurement, and objective analysis to document learnings over extended periods.
You should avoid relying on intuition for ambiguous problem solving when confirmation bias risks skewing decisions. Converting goals into an evidence-based decision loop with falsifiable tests ensures objective analysis and structured measurement guide your next actions.
Yes, goal-first success criteria require defining measurable indicators, thresholds, constraints, and anti-goals before running any tests. This ensures iteration and data analysis remain objective and aligned with falsifiable outcomes throughout the evidence loop.
musher-devGenerate problem frame documents from a context brief.
openscientist-ioRank hypotheses by impact, feasibility, and novelty to select the next research action.
ajbmachonRun hypothesis-test-measure-analyze-iterate cycles to converge on evidence-backed conclusions.
tutran27Decompose complex problems into revisable sequential thought sequences.
