figure-it-out

Designs an auditable phased playbook for large migrations and ambiguous multi-part changes.

2|Updated Jun 27, 2026
One-click install
npx skills add https://github.com/imjasonh/playground --skill figure-it-out-imjasonh
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: figure-it-out
Source: https://github.com/imjasonh/playground/tree/main/.cursor/skills/figure-it-out
Command: npx skills add https://github.com/imjasonh/playground --skill figure-it-out-imjasonh

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Ambitious tasks like large migrations or multi-part changes often have no existing playbook, leading to ad-hoc execution, lost decisions, and work a human cannot audit after stepping away. This Skill designs a rigorous, falsifiable workflow before any code is written. ## Core Features & Use Cases - Framing and rigor scaling: Defines done as a falsifiable predicate, quantifies scope, and assigns rigor levels based on blast radius and reversibility. - Hypothesis-driven execution loop: Runs each unit as an experiment with VERIFIED, NOT VERIFIED, or INCONCLUSIVE verdicts, verifying artifacts directly rather than trusting self-reports. - Auditable decision trail: Logs every decision and unit in a canonical TSV via the show-me-your-work skill so reviewers can trust the work from the PR. - Use Case: When asked to migrate a large codebase to a new framework, use this Skill to decompose the work into atomic landable units, build a verification harness from the pre-change baseline, and produce a committed decision log for human review. ## Quick Start Ask the AI to figure out a plan for migrating the legacy billing module to the new payments API with a verifiable audit trail.

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?▼

Frame the migration as a falsifiable definition of done, decompose it into atomic independently-landable units, and sequence the riskiest unknowns first. Build a verification harness from the pre-change baseline before starting feature work.

How to make autonomous AI coding work auditable for reviewers?▼

Log every decision and work unit in a canonical TSV file with evidence links, committed alongside the diff. The trail plus the diff lets a human return later and trust the work without re-running it.

When should I use a general playbook instead of a task-specific one?▼

Use a general playbook when the task matches no narrower workflow, such as an ambiguous multi-part change or a large migration. It scales rigor to the task, applying more gates and artifacts to one-way doors and high blast radius work.

What is a hypothesis loop in AI-driven development?▼

Each unit of work is treated as an experiment: state a hypothesis, make the smallest change, and measure against the predicate on the real artifact. Keep changes that advance the goal, revert those that do not, and never treat an inconclusive result as a pass.

When should parallel fan-out be avoided in multi-step tasks?▼

Parallelize only across clear seams, giving each worker its own worktree or branch to avoid serializing shared state. Over-fanning adds coordination cost, and a second design review over an already-settled design is over-engineering.