AGENT_AUDIT_LOGGING
Verify AI agent actions produce immutable audit logs with full provenance chains.
Framework Mappings
AIUC-1 E005 — LoggingSOC 2 CC7.2 — Logging & MonitoringOWASP Agentic ASI10 — Rogue Agents
Agent Frameworks Scanned
Claude CodeLangChainCrewAIAutoGenMCP
Evaluation Facets
01
Logging Middleware
RequiredWeight: 0.3
Pass:Middleware captures all agent actions, tool calls, model invocations
Fail:No logging middleware
02
Log Schema Completeness
RequiredWeight: 0.25
Pass:All required fields: action, agent identity, inputs, outputs, tools, timestamps
Partial:Some fields missing but basic logging present
Fail:Minimal or no structured schema
03
Immutability
Weight: 0.25
Pass:Append-only/write-once storage (S3 Object Lock, CloudWatch Logs)
Partial:Centralized but not immutable
Fail:Local mutable logs
04
Provenance Chain
Weight: 0.2
Pass:Full provenance with correlation IDs and trace/span linking
Partial:Basic logging without correlation
Fail:No provenance tracking
Remediation Steps
- 1Add logging middleware capturing all agent actions
- 2Define structured log schema with required fields
- 3Use immutable, append-only storage for audit logs
- 4Add correlation IDs linking actions across the chain
Evaluate this control automatically
Connect your repos and Brigs evaluates AGENT_AUDIT_LOGGING across all your agent configurations.
Get Started — Free