OUTCOMES
What this Prompt helps you deliver
- A repeatable acceptance scenario exists
- Failures identify the responsible layer
BEFORE YOU START
Before you start
- The user flow and expected result are explicit
- The customized feature runs in the target environment
What AI should inspect first
docs/agent/ai-harness-principles.mddocs/agent/electron-e2e-plan.mddocs/agent/dataflow-verification.mdappsREADY TO RUN
Complete Prompt
Copy the complete Prompt with its execution and acceptance boundaries.
Build acceptance that proves the TabTin user flow completes and distinguishes driver, API, persistence, realtime, and rendering failures—not a test that a button exists. Read AGENTS.md, docs/agent/ai-harness-principles.md, docs/agent/electron-e2e-plan.md, docs/agent/dataflow-verification.md, and relevant apps. Collect goal, initial state, actions, observable result, targets, and risk, then choose the closest stable real driver.
Write Given/When/Then with data, identity, environment, and success. Provide six primitives: drive, observe, judge, attribute, evidence, discover. Reuse real API, clients, and components; replace only slow external services at a clear boundary. Assert user-visible and persistence/event results. On failure retain minimal sufficient logs, screenshot, events, or DB evidence linked to the step.
Watch the test fail for the missing behavior or injected fault, then pass. Cover success, empty, permission failure, and critical dependency failure. Repeat without order dependence or fragile sleeps; compile mobile platforms and use PostgreSQL for backend tests. Report scenarios, levels, commands, evidence, attribution, pass results, uncovered devices/systems, and the product regression each test catches.