Skip to main content
v2 draft outline

Outline only. Bullets list what this page will eventually cover.

Module 2: Diagnose the Bad Repo

Sizing: 5 days (Week 2); first half of converting a bad repo to a good one

Outcome​

  • Use AI to compare the bad repo against the good one and find what is missing
  • Produce a prioritized plan to close the gaps

Framing example​

  • A non-engineer built a dashboard with an AI coding tool. It works. Now what?
  • What would you want to be true before this reaches real users?

The bad repo: Chaos Emerald Hotels​

  • Repo: liatrio-labs/chaos-emerald-hotels, the degraded sibling of Emerald Grove Hotels
  • Same app lineage as the good repo, so the comparison is fair
  • It works on the happy path; most of the factory is missing, and some tooling is present but not enforced
  • Seeded flaws were planted on purpose
  • The inventory of flaws is instructor-only; learners find them
  • Do not link learner material to the instructor key or its pull request

What you'll do​

  • Explore Chaos with an agent
  • Compare it to the Module 1 good repo, piece by piece
  • List what is missing and what is present but ignored
  • Rank the gaps by risk and cost to close
  • Write the plan Module 3 will execute
  • Include any AGENTS.md and standards you would need to write down

Using AI to compare​

  • Ask for a gap report, then check it yourself against the good repo
  • Keep the report short enough to review; push back anything unparseable
  • Note what the agent missed or got wrong
  • Facilitators compare each learner's list to the instructor key afterwards

Gate​

  • No new gate this week; diagnosis and planning
  • Bug fixes and small features are just work that will pass through the factory later, not the point

Anti-patterns​

  • Fixing things as you find them with no plan
  • Trusting the agent's gap report unchecked
  • Treating a wired-in scanner as working when its findings are ignored
  • Rubber-stamping a wall of agent text

Demo​

  • Process, not fixes: how you compared, what the agent found, what you ranked first and why
  • Defend your plan: explain each priority if questioned

From v1​

  • Replaces the old Foundations module
  • Reuses the by-hand and skills ideas only where they help diagnose

TODO​

  • TODO: confirm each learner forks Chaos, as in Week 1, rather than sharing one copy
  • TODO: whether the readiness audit tool is used for the comparison, and what starting score Chaos should show
  • TODO: what the deliverable plan looks like (format and length)
  • TODO: how facilitators grade a learner's gap list against the instructor key without leaking it
  • TODO: which seeded material from the earlier draft is reused (parked on the overview)