Purpose: Agent orchestration architecture, RAG design, prompt management, model routing/cost controls, AI evaluation strategy, red-teaming, and the model risk register.
What belongs here: Design and governance documentation for AI capabilities. Actual prompt text lives in prompts/; evaluation datasets/cases live in evals/.
What must not be stored here: Real candidate data used as examples, unversioned prompt changes, or production model API keys.
Owner: [TENANT_CONFIGURATION_REQUIRED — AI Governance Lead]
Main dependencies: prompts/, evals/, config/schemas/rag-config.schema.json, config/schemas/model-routing.schema.json, .claude/rules/ai-agents.md.
Contents: agent-architecture.md · agent-skill-catalog.md · agent-state-and-orchestration.md · rag-architecture.md · rag-ingestion-and-chunking.md · retrieval-and-grounding-policy.md · model-routing-and-cost-controls.md · prompt-management.md · ai-evaluation-strategy.md · red-team-plan.md · model-risk-register.md
Status: Initial scaffold — details to be added during implementation.