figure-it-out

Designs an auditable multi-phase playbook for large migrations and ambiguous engineering tasks.

Updated Sep 2, 2026
One-click install
npx skills add https://github.com/jnyross/pstack-muse --skill figure-it-out-jnyross
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: figure-it-out
Source: https://github.com/jnyross/pstack-muse/tree/main/skills/figure-it-out
Command: npx skills add https://github.com/jnyross/pstack-muse --skill figure-it-out-jnyross

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a task matches no existing playbook—a large migration, an ambitious multi-part change, or work a human reviews after stepping away—this Skill designs a rigorous, auditable workflow before any code is written, so the effort doesn't drift or produce unverifiable results. ## Core Features & Use Cases - Playbook Design: Frames the task with a falsifiable definition of done, quantified scope, and a rigor level scaled to risk, then decomposes the work into atomic, independently-landable units. - Hypothesis Loop: Runs each unit as an experiment—state a hypothesis, make the smallest change, verify against the predicate on the real artifact, keep or revert—with VERIFIED, NOT VERIFIED, or INCONCLUSIVE verdicts. - Audit Trail: Logs every decision and unit in a canonical TSV via the show-me-your-work skill, with evidence links a reviewer can re-run, so a human can trust the work after stepping away. - Use Case: Migrating an API across dozens of call sites? The Skill sequences riskiest-unknown-first, builds the verification harness before the changes, and commits a decision trail alongside the diff for PR review. ## Quick Start Ask the agent to figure out a plan for migrating the payment module to the new API, with a decision log I can review afterward.

Frequently Asked Questions about figure-it-out

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I plan a large code migration with an AI agent?

Use a playbook-design workflow: frame a falsifiable definition of done, quantify scope and blockers, decompose into atomic independently-landable units, and sequence riskiest-unknown-first. Build the verification harness before making changes so each unit is checked against a baseline.

How to make autonomous AI coding work auditable?

Keep a canonical decision log—one TSV row per decision and per unit, with evidence as links produced by committed scripts a reviewer can re-run. Commit the trail alongside the diff so the human reviewing the PR can verify what was decided and why.

When should I use figure-it-out instead of a bug fix or feature playbook?

Use it when the task matches no narrower playbook: a migration across many call sites, an ambitious multi-part change, or work reviewed after stepping away. A focused single-unit task that fits Bug fix, Feature, or Perf should route to those instead.

What is a hypothesis loop in AI-assisted engineering?

Each unit of work is treated as an experiment: state the hypothesis, make the smallest change, measure against the predicate on the real artifact, keep it if it advanced, revert if not. Verdicts are VERIFIED, NOT VERIFIED, or INCONCLUSIVE—inconclusive is not a pass.

When should design decisions not go through an arena-style review?

Skip multi-candidate design review for mechanical work whose shape is already concrete. Running a second arena over a settled design is over-engineering; reserve it for one-way-door decisions with high blast radius.