SovAIHub
ModulesSAI-230
SAI-230 table of contents
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

  1. Why must permission enforcement happen before ranking and generation?
  2. Which lineage fields let a citation resolve to an exact source?
  3. How should retrieved prompt-injection text be treated?
  4. What is the difference between retrieval relevance and answer groundedness?
  5. When should the system produce no answer?
  6. Why must deletion tests cover caches and evaluation data?
  7. Which changes can invalidate a prior RAG evaluation?
  8. 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.