test-scenarioslisted

# Test Scenarios Turns a spec or story into a complete test matrix using classic test-case design: happy path, edge cases, error paths, and data variations — plus quality-bar/eval cases for ML and AI features where output is probabilistic rather than deterministic. **Grounded in:** *Agile Testing* — Lisa Crispin & Janet Gregory: cover happy/edge/error/data paths from acceptance criteria. **Go deeper (The Product Channel):** [PM's 30-Minute AI Evals](https://sidsaladi.substack.com/p/pms-30-minute-ai-evals-a-lightweight) ## When to use this - A spec or story is "code complete" and you need acceptance test cases before sign-off. - Engineering says "it works" but you suspect untested edge and error paths. - You're writing acceptance criteria and want them backed by concrete, checkable scenarios. - A feature involves messy real-world data (uploads, dates, currencies, names, timezones) and you need data-variation coverage. - The feature is ML/AI-driven (search, recommendations, classification, generation) and you need eval cases with a defined quality bar, not just pass/fail. ## Before you start (gather these) - **The spec or story** — what the feature does, the user goal, and the acceptance criteria as written. - **Inputs and their constraints** — every field/parameter, allowed types, ranges, required vs. optional, formats. - **System boundaries** — dependencies, integrations, auth/permission rules, and what's explicitly out of scope. - **For ML/AI features** — what "good outp