Deterministic · Air-Gapped · Zero-Egress · FIPS 140-3

Deterministic AI Governance for Federal Contractors.

PromptFrame produces the governance evidence federal AI deployment requires — deterministic scoring against federal frameworks and the complete AI-governance evidence package for ATO preparation (the AI-system portion of the body of evidence), including machine-readable OSCAL that feeds your authorization platform. No LLM in the scoring path: same input, same result, every time. Self-hosted on your infrastructure. A complementary runtime enforcement layer is on the roadmap.

Request Access See the Platform
Frameworks: NIST AI RMF EO 14179 OMB M-25-21/22/26-04/M-26-14 GSAR 552.239-7001 (proposed) CMMC Level 2 NIST SP 800-53 PA EO 2023-19

The Platform

Design-Time Governance. The Product.

PromptFrame's design-time platform scores AI systems and generates the ATO evidence — deterministic, air-gapped, machine-readable. A complementary runtime enforcement layer is on the roadmap; the shipping product is design-time.

Design-Time (DT)
Governance Scoring & ATO Artifact Generation

Deterministic 10-dimension scoring of AI system prompts. No LLM in the scoring path — same input always produces same output. The scoring method is defensible under C3PAO and 3PAO scrutiny. Auto-generates the complete AI-governance evidence package for ATO preparation (the AI-system portion of the body of evidence) per assessment.

  • SSP narratives (per dimension)
  • NIST SP 800-53 Rev 5 control family crosswalk
  • POA&M in FedRAMP format
  • GSAR 552.239-7001 (proposed) compliance checklist (14 paragraphs)
  • SPRS export
  • Per-dimension remediation report
  • Executive engagement summary
  • All artifacts SHA-256 integrity-protected · HMAC-signed audit chain
  • Auditor submits · Approver locks — enforced at database level (separation of duties)
  • 15 built-in use case templates (internal assistant, agentic workflow, GovCon proposal AI, code generation, decision support, and more)
Runtime (RT) · Roadmap
Policy Enforcement & Audit Trail — In Development

A complementary runtime enforcement layer, designed to sit inline with LLM and agentic toolchains so every tool call executes within the approved policy boundary, each decision a cryptographically signed audit record. Architected and in development; not yet fielded. The shipping product today is Design-Time.

  • Tool authorization — designed so only approved tools execute
  • Scope boundary — designed to keep an agent within its defined purpose
  • Data handling — planned tool-sequence and scope policy to gate unauthorized egress paths
  • Privilege — designed to gate escalation beyond authorized access
  • Designed to promote runtime findings back to the Design-Time audit chain
  • Deterministic by design — same policy, same decision
  • Model-agnostic — gates tool-calls from local air-gapped models routed through the gate
  • Status: designed, in development, not yet fielded.
Shadow AI Discovery
Surface AI Systems You Don't Know You Have

Standalone scanner for Windows, macOS, and Linux. Detects installed AI applications, browser extensions, IDE plugins, local model runners (Ollama, LM Studio), MCP server configs, and API credential files. Analyzes network logs in six formats: Apache/Nginx CLF, ArcSight CEF, CSV, DNS query logs, Cisco ASA/FTD syslog, and directory scan output.

  • Foreign-origin AI vendor flagging per GSAR 552.239-7001 §(e)(2) (proposed rule)
  • DeepSeek (China) · Mistral (France) classified
  • Classifies discovered systems by governance type
  • Cryptographically signed scan report
  • Importable into governance record as chain-of-custody artifact
  • Satisfies OMB M-26-14 CEM requirements at AI egress layer
10
DT Governance Dimensions
4
RT Enforcement Categories
14
GSAR Paragraphs Mapped
FIPS 140-3
Government-Grade Encryption
Cryptographically Signed
Zero External API Calls
CMMC Level 2 — Full Module
110 Practices · 14 Domains
Weighted SPRS score per DoD Assessment Methodology v1.2.1. Gap flags auto-generate POA&M items. Assessment findings tracked through workflow state machine to PIEE export. Produces the documentation evidence C3PAO assessors require.
OMB M-26-14
CEM & THIRF
Compliance
Shadow AI scanner satisfies the Continuous Evaluation Mechanism (CEM) at the AI egress layer and the Threat and Harm Incident Reporting Framework (THIRF) requirements. 5-level maturity model mapped. Signed evidence artifact per assessment — importable into your M-26-14 compliance record.

Architecture

The Continuous Governance Loop

A point-in-time audit cannot satisfy a continuously evolving AI deployment. The design-time platform delivers the design → approve → attest → decay → re-approve cycle today; the enforce step and its real-time feedback are part of the roadmap runtime layer. The full loop keeps your governance posture current as AI systems change.

📋
Design
Score AI prompt across 10 dimensions
🔐
Approve
Separate approver locks policy — dual-role enforced
🛡️
Enforce
Roadmap RT gate: block unauthorized tool calls inline
Attest
Design-time re-attestation advances NIST control status (roadmap: runtime evidence)
⏱️
Decay
Controls regress after 90 days without attestation
🔄
Re-Approve
Changed posture requires a new approval record
90-day attestation TTL — Controls that go un-attested automatically regress, forcing re-engagement. An organization cannot certify once and walk away.
Runtime findings flow back (roadmap) — The design is for every runtime gate decision and anomaly to promote to the design-time audit chain, updating posture continuously. This closed loop is part of the roadmap runtime layer, not the shipping product.

Why Continuous Governance

The Math Behind the Model

This is not a design philosophy. It is a mathematical requirement derived from peer-reviewed research at NIST.

In a paper published in IEEE Security & Privacy, Apostol Vassilev of NIST applied Gödel's incompleteness theorems — mathematical results from 1931 proving the inherent limits of any formal rule system — to the domain of AI security.

The proof concerns adversarial robustness — whether a fixed rule set can withstand attack — and NIST draws an architectural conclusion from it: organizations must move from “one and done” security to continuous monitoring, with red-teaming, ongoing guardrail updates, and resilience planning for when exploits occur. The objective NIST states is making exploitation economically prohibitive, not achieving perfect impermeability. The implication for governance is that a point-in-time artifact cannot be the whole answer. PromptFrame delivers the design-time and re-attestation portions of that model today; the continuous-enforcement portion is the roadmap runtime layer.

This is why design-time governance is necessary but not sufficient on its own. A single compliance report, a point-in-time audit, or a static policy document answers what a system was designed to do — not how it behaves under adversarial pressure over time. Both layers are required.

"What this proof shows is that there is no finite set of guardrails that is universally robust against adversarial prompts."

Vassilev, A., “Robust AI Security and Alignment: A Sisyphean Endeavor?” — IEEE Security & Privacy, May 2026 · NIST
PromptFrame's Answer
Approval cycle — every prompt version scored and locked before deployment
Runtime enforcement (roadmap) — designed to gate every tool call against approved policy
Control attestation (roadmap) — runtime evidence designed to advance NIST control status
TTL decay — status regresses if not re-attested, forcing re-engagement
Re-approval — changed posture requires a new signed approval record

* GSAR 552.239-7001 is a proposed rule pending GSA finalization. GSA published a Revised Clause on June 17, 2026; the public comment period closes August 3, 2026 (public listening session July 14, 2026). No enactment date has been announced. GotHawk submitted a formal public comment April 3, 2026, recommending GSA add a design-time documentation requirement — the governance gap PromptFrame was built to close.

Assessment Output

What DT Produces

Every DT assessment produces a complete, SHA-256 integrity-protected artifact package — deterministic, evidence-based, ready for a contracting officer, AO, or C3PAO assessor. Output includes machine-readable OSCAL (SSP + POA&M) so the evidence feeds an OSCAL-native authorization platform, not just a PDF.

32
Overall Governance Score (0–100)
Non-Compliant
Example: Ollama local model · 10 gaps · 6 critical
SHA-256 verified · HMAC-signed audit chain
Compliance Tier Non-Compliant (<70)
GovCon Ready No — critical gaps
Dimensions Scored 10 of 10
ATO Artifacts Generated 7 — all SHA-256 signed
NIST 800-53 Crosswalk Included
POA&M (FedRAMP format) Included
GSAR Checklist (proposed rule) Included

Scoring Framework

10 Design-Time Governance Dimensions

Every AI system prompt is scored 0–10 across all 10 dimensions. Three dimensions are critical. A security dimension score below 3 triggers a GovCon-Not-Ready override regardless of overall score.

Dimension 01
Prompt Clarity
NIST AI RMF 1.0 GOVERN 1.1 · OMB M-25-21 §3 · OMB M-25-22 §3
Dimension 02
Agentic Architecture
NIST AI RMF 1.0 MANAGE 1.3 · OMB M-25-21 §4 · OMB M-25-21 §5
Dimension 03
Tool Boundary Definition
NIST AI RMF 1.0 MEASURE 2.5 · OMB M-25-21 §3 · GSAR 552.239-7001 (proposed) §(e)(1)
Dimension 04
Data Privacy & Security
NIST AI RMF 1.0 MANAGE 2.2 · OMB M-25-21 §4 · GSAR 552.239-7001 (proposed) §(d)(3)(i)
Critical
Dimension 05
Transparency & Disclosure
NIST AI RMF 1.0 GOVERN 4.1 · OMB M-26-04 §3 · OMB M-25-21 §4
Dimension 06
Policy Adherence
NIST AI RMF 1.0 GOVERN 1.2 · OMB M-25-21 §4 · OMB M-25-22 §3
Critical
Dimension 07
Output Quality Controls
NIST AI RMF 1.0 MEASURE 2.5 · OMB M-25-21 §4 · OMB M-26-04 §2
Dimension 08
Context & State Integrity
NIST AI RMF 1.0 MANAGE 2.2 · OMB M-25-21 §4 · GSAR 552.239-7001 (proposed) §(d)(3)(i)
Dimension 09
Explainability
NIST AI RMF 1.0 GOVERN 1.1 · OMB M-26-04 §2 · GSAR 552.239-7001 (proposed) §(e)(3)
Critical
Dimension 10
Task Alignment
NIST AI RMF 1.0 GOVERN 1.1 · OMB M-25-21 §3 · OMB M-25-22 §3

Deployment

Self-Hosted. Your Infrastructure. Your Data.

PromptFrame runs on your infrastructure as a Docker container stack. GotHawk delivers signed container images — no data is ever transmitted to GotHawk or any third party.

Self-Hosted Container
Available Now

Your team runs the Docker stack on your own servers or cloud infrastructure. Compatible with air-gapped networks and CUI environments. FIPS 140-3 capable.

✓  Signed Docker container images
✓  Deployment documentation
✓  CUI / air-gapped environment configuration
✓  SHA-256 pinned framework file
✓  Ongoing image updates and security patches
Data Handling Guarantee

GotHawk never receives, processes, stores, or trains on client prompt data or assessment outputs. All data stays within your infrastructure boundary.

Security Specs
FIPS 140-3 capable
Government-grade encryption at rest and in transit
Cryptographically signed audit chain
Tamper-evident · independently verifiable
Zero external API calls

Intended Users

Built For

Any organization deploying AI systems that needs defensible governance documentation — from small contractors to defense primes.

🏢
Small Federal Contractors

Document AI system compliance for use-case inventories and procurement packages without a dedicated GRC team. Self-hosted deployment means no external data exposure.

🏛️
Defense Primes & Mid-Tier Contractors

Vet AI components across your delivery environment. Establish a documented compliance baseline for AI systems in contract performance. CMMC Level 2 and FedRAMP Moderate posture support.

🛡️
DoD Program Offices

Accelerate compliant agentic AI deployment by producing governance evidence up front. The roadmap runtime layer is designed to keep every deployed agent within its approved policy boundary, with a signed audit trail for oversight review.

⚖️
Compliance & Risk Teams

Build AI use-case inventories and governance documentation for CMMC pre-assessments and ATO preparation. Deterministic scoring produces independently verifiable evidence.

🌿
PA State Agencies & Vendors

PromptFrame maps to PA EO 2023-19. GotHawk is a PA-registered small business, BDISBO self-certified, active in the PA Suppliers Portal (Jaggaer). Engagements available through the portal or direct award.

🧠
AI Product Teams

Score system prompts before deployment to identify governance gaps early — before procurement reviews or buyer due diligence surfaces them in a contract.


Deploy PromptFrame

Contact Williams Hawkins III to discuss your deployment. We'll walk through your AI systems, delivery environment, and compliance requirements — and get you set up with signed container images and deployment documentation.

Request Access

717-489-9585  ·  williams@gothawksolutionsllc.com  ·  We respond within one business day


Pennsylvania State Procurement

Available to PA State Agencies & Contractors

State agencies and Pennsylvania-registered contractors deploying AI systems face the same governance gap as federal buyers. PromptFrame maps to PA EO 2023-19 and produces documented, reviewable evidence of AI system configuration for IT risk reviews, vendor due diligence, and internal AI governance policies.

GotHawk Solutions LLC is a Pennsylvania-based small business, BDISBO self-certified, and active in the PA Suppliers Portal. Engagements can be structured through the portal or via direct award under applicable thresholds.

BDISBO Small Business Self-Certified BDISBO Micro Business Self-Certified PA Suppliers Portal Active (Jaggaer) Vendor No. 0000569874 Commodity 43230000 Commodity 80100000 Commodity 81111800 Commodity 84110000
Contact Us for a PA State Engagement PA Vendor Credentials → Capability Statement →