Assessment1 min readPrototype
SAI-230 knowledge check
Verify ingestion, lineage, permissions, retrieval, citation, grounding, no-answer, and evidence behavior.
Last content review 2026-08-03Included in SAI-230
Scenario questions
- Why must permission enforcement happen before ranking and generation?
- Which lineage fields let a citation resolve to an exact source?
- How should retrieved prompt-injection text be treated?
- What is the difference between retrieval relevance and answer groundedness?
- When should the system produce no answer?
- Why must deletion tests cover caches and evaluation data?
- Which changes can invalidate a prior RAG evaluation?
- Why should severe permission failures not be averaged into one score?
Completion criteria
The design passes when source authority, lineage, permissions, retrieval, citations, supportedness, no-answer behavior, threat cases, release gates, change triggers, telemetry, and deletion are explicit and tested with representative evidence.