← Back to the case study

For Technical & AI

Testing & acceptance

Behavioral checks for security, recovery, evidence, and side effects.

Public reading guide · Source: docs/09-Testing-Acceptance.md
A reviewed summary of the repository documentation and current implementation. Private operational details and phase-test records are omitted.

Core checks

Tests cover upload boundaries, page chunking, optional and estimated dates, recurrence arithmetic, partial failure, append/delete behavior, and cross-workspace access rejection. Calendar checks verify eligibility and parser-valid all-day artifacts.

Agent and workflow checks

Email checks cover webhook replay, identity, MCP scope, pause/resume, stale approvals, and duplicate sends. RAG checks cover workspace retrieval, citations, unsupported answers, index reuse, and deletion. QA checks seed invalid dates, unsupported evidence, and incorrect calendar artifacts.

Evidence and limits

Automated checks use mocks and synthetic fixtures. Separate local smoke tests establish provider integration where explicitly authorized. Passing tests do not establish universal AI accuracy or measured product outcomes. Each phase requires the owner's acceptance.

Explore the interactive architecture →