System area
Agentic Systems
Bounded tools, orchestration, evaluation, and inspectable traces for systems that can act.
Scope
Where it helps
Use this area when a workflow needs tool use, delegation, recovery, or accountable autonomy.
Inquiry
Questions LucyAI asks
- What authority does each tool require?
- Which traces make decisions inspectable?
- How do failures stop, recover, or escalate?
Outputs
Artifact ledger
- Tool-boundary map
- Agent workflow contract
- Evaluation and trace plan
Boundaries
Control ledger
- Least-privilege tool access
- Explicit stop conditions
- Recorded evaluation scope
Evidence
Source basis
- Building effective agents
Agent and workflow design practice
- Demystifying evals for AI agents
Evaluation vocabulary and practice
- NIST Generative AI Profile
Voluntary risk-management framing