Governance & Compliance · Software component

Policy Adherence Evaluator

Software componentGovernance & ComplianceSafety, Security & Governancearc:PolicyAdherenceEvaluator

An evaluation component that checks an agent's actions against domain policy documents and regulatory constraints, scoring task completion jointly with policy compliance.

Responsibility. Detects policy violations in agent runs and computes policy-aware success.

Also known as: Policy-aware evaluation, Automated compliance checking, Online Action Validator, Post-hoc Action Audit, Action Sequence Auditor

Direct neighbourhood (hover for relationship types)

Relationships

is configured by structural

Action Sequence Policy Ch3.8

is invoked by dependency

reads dependency

sends data to dynamic

audits assurance

Agent Controller abstract Ch3.8
Web Navigation Agent Ch3.3

evaluates assurance

Agent Controller abstract Ch9.2

Design guidance

MUST treat policy violations as failures regardless of task completion in regulated domains.
SHOULD give policy compliance equal weight to task completion in evaluation.
SHOULD run in production as automated compliance checking, not only pre-deployment.
SHOULD validate production actions against policies rather than reference trajectories.
SHOULD report near-misses prevented by validation layers to inform prompt refinement and retraining.

Quantitative guidance

As stated by the sources; verify before use.

Under joint scoring, 95% task success with a 10% violation rate scores below 90% success with zero violations (Ch3.3).

Classification

Patterns: Completion Under Policies (CuP)Joint task-and-policy metricOnline policy-level action validationLoop detection (same tool, same parameters after failure)Weekly post-hoc safety audit with near-miss reporting
Technologies: ST-WebAgentBenchtau-bench
Quality attributes: Transparency and accountability (NIST AI RMF: accountable and transparent)
Risks mitigated: Tasks completed through policy violationsRegulatory liabilityProcedural violations invisible to parameter-level checksRetry loopsProhibited or destructive tool use

Sources

Ch3.3: T. Nguyen, "Web Navigation and Interaction Benchmarks," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.3. ISBN: 9798244538229.
Ch3.8: T. Nguyen, "Action Accuracy Metrics," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.8. ISBN: 9798244538229.
Ch9.2: T. Nguyen, "Action Constraints and Permission Models," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 9.2. ISBN: 9798244538229.