Playbook

Scenarios (static MVP)

Each scenario is a structured workflow: what to gather, what to ask, and in what order—so AI answers stay useful in messy real systems.

• Slow or flaky API

  Trace latency through layers: gateway, app, DB, cache. Prompts aligned to each signal you can pull.

  Debugging

• Feature from issue to PR

  Scope, contracts, edge cases, then implementation slices—each step has prompt patterns tied to review checkpoints.

  Build

• Production incident triage

  Time-boxed path: customer impact → recent deploys → error budget → targeted log queries and rollback decisions.

  Prod

• Test failures after a refactor

  Diff-first narrowing: failing test → changed assumptions → minimal repro prompts.

  Debugging

• Onboarding to an unfamiliar codebase

  Map ownership, entry points, and data flow before asking “explain this repo.”

  Investigate