SW
Agent governance

Every agent is trained by a named auditor.

This is specialist accountability, surfaced as a product feature. Each framework agent has a designated Trained by field — the credentialed SME whose review methodology, finding language, and evidence standards the agent absorbs through the redline loop. Prompt traces, source references, and approval gates stay visible so the auditor can see where AI helped and where a human signed off.

Agents in production
12
SMEs assigned
1 / 12
Avg calibration
81%
Redline loop
Human override rate
1.8%
the auditor still decides
Filter
SOC 2 Auditor Agent
SOC 2 · auditor
SME TBD

Role-plays a senior AICPA SOC practitioner. Challenges weak evidence, cites 2017 TSC with 2022 Points of Focus, and raises the concerns a real practitioner raises in fieldwork.

Trained by
— awaiting workspace SME —
AICPA SOC Practitioner
Calibration
78%
Artifacts
1,242
Findings
87
2026-04-22override rate · 1.13%
Specialty: CC6.1–CC8.1 · logical access + system operations
HITRUST r2 Assessor Agent
HITRUST r2 · auditor
SME assigned

Role-plays a HITRUST CCSFP performing a pre-assessment walkthrough. Preserves native references and flags assessor evidence gaps without copying proprietary requirement text.

Trained by
named
HITRUST Reviewer
HITRUST CSF Assessor (CCSFP)
Calibration
91%
Artifacts
2,014
Findings
131
2026-04-23override rate · 0.45%
Specialty: 01.b–01.v · access control family + 11.a incident reporting
CMS EDE Auditor Agent
CMS EDE · auditor
SME TBD

Role-plays a CMS EDE third-party auditor performing a Year 9 Operational Readiness Review. Cites guidance by name and date.

Trained by
— awaiting workspace SME —
CMS EDE Third-party Auditor
Calibration
73%
Artifacts
518
Findings
62
2026-04-22override rate · 4.05%
Specialty: Phase 1 / 2 / 3 ORR + MARS-E 2.2 + Privacy Act
HIPAA Security Rule Reviewer
HIPAA · auditor
SME TBD

Reviews ePHI safeguards against 45 CFR § 164 Subpart C. Flags risk-analysis gaps per 164.308(a)(1)(ii)(A) and HHS OCR enforcement patterns.

Trained by
— awaiting workspace SME —
CHSPP
Calibration
84%
Artifacts
1,680
Findings
74
2026-04-21override rate · 0.65%
Specialty: 164.308 Administrative + 164.312 Technical Safeguards
PCI DSS 4.0.1 QSA Agent
PCI 4.0.1 · auditor
SME TBD

Writes findings to the v4.0.1 Report on Compliance template. Tracks future-dated requirements (effective 2025-03-31).

Trained by
— awaiting workspace SME —
PCI QSA
Calibration
69%
Artifacts
392
Findings
41
2026-04-20override rate · 4.59%
Specialty: Req 3 PAN protection · Req 8 MFA · Req 10 logging · Req 11 scanning
ISO 42001 Lead Auditor Agent
ISO 42001 · auditor
SME TBD

Reviews AI Management System clauses against ISO/IEC 42001:2023. Integrates NIST AI RMF (AI 100-1) and Generative AI Profile (AI 600-1) references.

Trained by
— awaiting workspace SME —
ISO 42001 Lead Auditor / CAISS
Calibration
76%
Artifacts
264
Findings
29
2026-04-22override rate · 3.03%
Specialty: A.6.1 AI impact assessment · A.7.2 data quality · A.9.2 responsible use
Cross-Framework Mapping Agent
Cross-FW · mapping
SME TBD

Reads one evidence artifact and produces structured cross-framework control mappings with calibrated confidence, evidence type, and auditor-grade rationale.

Trained by
— awaiting workspace SME —
Framework SME Panel
Calibration
88%
Artifacts
4,421
Findings
0
2026-04-23override rate · 1.85%
Specialty: Calibrated confidence + evidence typing
Source Library Ingestion Agent
Intake · ingestion
SME TBD

Classifies, dedupes, and extracts requirements from a client's local source folder. Metadata-only (filenames, sizes, timestamps) — contents never leave the browser.

Trained by
— awaiting workspace SME —
CMS zONE-aware SME
Calibration
82%
Artifacts
1,029
Findings
0
2026-04-23override rate · 3.3%
Specialty: CMS EDE Year 9 source taxonomy
RAG Artifact Generator
RAG · generator
SME TBD

Drafts policies, procedures, SSP sections, POA&M entries, and control narratives grounded in the client corpus. Inline citation chips. Refuses to generate without source anchor.

Trained by
— awaiting workspace SME —
House-style SME
Calibration
85%
Artifacts
631
Findings
0
2026-04-23override rate · 7.45%
Specialty: Auditor-accept score loop (Karpathy)
Gap Analyzer
Gap · gap
SME TBD

Identifies missing, stale, or weak evidence per framework. Proposes a prioritized remediation plan that maximizes cross-framework reuse.

Trained by
— awaiting workspace SME —
Engagement lead SME
Calibration
80%
Artifacts
892
Findings
146
2026-04-23override rate · 2.47%
Specialty: Stale/weak/missing evidence triage
Control Testing Agent
Test · testing
SME TBD

Executes auditor-style test procedures against the client's evidence — identifies exactly where the flow fails with cited artifacts and user-level specificity.

Trained by
— awaiting workspace SME —
Senior test-lead SME
Calibration
74%
Artifacts
156
Findings
38
2026-04-23override rate · 12.18%
Specialty: Per-step pass/partial/fail with evidence citations
Evidence Reviewer Agent
Review · evidence review
SME TBD

Objective, dispassionate summary of the evidence state per framework. Feeds the Auditor agents for the pre-audit simulation.

Trained by
— awaiting workspace SME —
Framework SME (any)
Calibration
87%
Artifacts
3,114
Findings
0
2026-04-23override rate · 0.48%
Specialty: Evidence typing + reuse signal
How the training works

Each ControlFrame agent runs the redline loop: Propose → Act → Evaluate → Reflect → Refine. The named SME reviews a draft, marks their changes, and the agent captures the pattern into a firm-tunable retrieval corpus. Over time, the agent's drafts approach the SME's auditor-accept threshold — without ever replacing the SME's signature. Judgment, sign-off, and accountability stay human.

Agent Registry | ControlFrame