What problem does it solve?
It replaces unstructured brainstorming and gut-feel decisions with a rigorous First Principles Framework (FPF) cycle that generates competing hypotheses, verifies their logic, validates evidence, audits trust, and produces a documented decision.
Core Features & Use Cases
- Structured Hypothesis Pipeline: Generates L0 hypotheses, then promotes them through L1 (logic-verified) and L2 (evidence-validated) stages using parallel sub-agents, rejecting invalid ones.
- Trust Auditing & Decision Records: Audits each validated hypothesis with an R_eff score and weakest-link analysis, then writes a Design Rationale Record to
.fpf/decisions/.
- Interactive Refinement: Lets users add their own hypotheses and challenge the final decision, with persistent artifacts stored in a
.fpf/ directory for later review via /fpf:status.
- Use Case: When facing an ambiguous architecture or product problem, run the workflow to get competing hypotheses systematically filtered by logic and evidence, ending with an auditable decision record.
Quick Start
Run the propose-hypotheses skill with a problem statement such as "Should we migrate our monolith to microservices?" to generate, validate, and decide between competing hypotheses.
