Skip to content

Lift 2: Define the Work

Where We're Starting

After your gear check, you've got the fundamentals: how AI behaves, how to write clear prompts with the Three Pillars, how to create delegation contracts with user stories, and how to give AI standing instructions through your project context file. You put those tools to work in Run 1, building the core of your Avalanche Observation Network.

But something probably started to bug you. You found yourself repeating the same instructions — how to add a page, how to structure a form, how to break down a feature — and getting slightly different results every time. The processes you followed in one conversation didn't carry over to the next. And the backlog of things you want to build is bigger than any single story can capture.

There has to be a more reliable way. There is. This lift gives you three tools to take control: decomposition that breaks big goals into delegatable pieces, skills that encode your team's processes so AI follows them consistently, and manual review that checks output against your acceptance criteria before anything ships.

What You'll Learn

  • How to break big goals into collections of independently shippable stories you can delegate and track
  • How to capture your team's repeatable processes as reusable AI instructions (skills) that work the same way every time
  • How to manually verify AI output against your acceptance criteria — the quality gate before anything ships

Sections

  1. Managing a Body of Work — Breaking big goals into delegatable, story-sized pieces
  2. Skills: Encoding Your Judgment — Reusable AI instructions that solve the consistency problem
  3. Manual Review: Your Eyes Are the Quality Gate — Checking AI output against acceptance criteria — pass or fail

By the End of This Lift

  • You can decompose a large feature into independently shippable story-sized pieces
  • You can explain what a skill is and why it solves the "different results every time" problem
  • You've built a skill using the "We Do, You Do" pattern — capturing a process you practiced as reusable instructions
  • You can review AI output against acceptance criteria and make a clear pass/fail call
  • You're ready to start Run 2 with a decomposed backlog, skills for consistency, and a review process