Managed Assessments

For organizations that need expert-led security evaluation, TrustTrace offers comprehensive OWASP LLM Top 10 assessments with live testing, compliance mapping, and a detailed remediation roadmap.

What You Get

A TrustTrace managed assessment is a 1–2 week engagement that evaluates your entire AI agent ecosystem. It includes everything the self-service scan covers, plus live adversarial testing, manual expert analysis, and a written report you can present to your board or auditors.

Assessment Scope

Assessment Tiers

How It Works

Before the Engagement

Passive Reconnaissance (Day 0) — Before we even speak, we run our passive recon scanner against your domain. We discover publicly accessible MCP servers, check for exposed API keys in public repositories, enumerate AI-related subdomains, and assess your vendors' compliance posture. You'll see these results in the kickoff call.

Day 1 — Kickoff

We conduct a structured intake interview with your technical team to inventory your AI agents. For each agent, we document: framework, tools, data access, PHI handling, MCP server connections, human-in-the-loop controls, and vendor BAA status.

This interview alone produces 30–40+ initial findings based on architectural risks — before any scanning begins.

Days 2–7 — Scanning and Analysis

As you provide materials (log exports, code access, tool schemas, MCP configurations), we run our full scanner suite:

• Log Scanner — Analyzes agent log files for PHI in plaintext, leaked secrets, system prompt fragments in error responses, and excessive data in tool call outputs
• Repo Scanner — Reviews agent source code for hardcoded credentials, prompt injection sinks, unsafe code execution, SQL injection vectors, and MCP-specific vulnerabilities
• Dependency Scanner — Checks all dependencies against CVE databases, identifies unpinned packages, detects typosquatting candidates
• MCP Scanner — Evaluates every MCP server your agents connect to: tool poisoning, authentication, transport security, rug pull risk, scope analysis
• Tool Auditor — Assesses every tool permission for least-privilege violations, dangerous patterns, and missing constraints

Our team also performs manual analysis that automated tools can't replicate: reading system prompts for subtle misconfigurations, cross-referencing findings with your specific compliance requirements, and identifying architectural risks unique to your deployment.

Days 5–10 — Live Testing (Enterprise Tier)

For Enterprise engagements, we run adversarial testing against your test or staging environments:

• Prompt injection attacks (direct and indirect)
• System prompt extraction attempts
• Data exfiltration probes
• Tool abuse scenarios
• MCP tool poisoning simulations

Every attack attempt is recorded with the exact payload, the agent's response, and an assessment of whether the attack succeeded. These are Confirmed findings — the highest confidence level.

Days 8–10 — Report and Readout

We generate a comprehensive assessment report and present findings in a readout call with your security and engineering leadership.

Report contents:

1. Executive Summary (1 page — designed for CISO/board consumption)
2. Engagement Overview
3. Overall Risk Score with OWASP Radar Chart
4. Critical and High Findings (detailed)
5. Medium and Low Findings (catalog)
6. MCP Security Assessment
7. Supply Chain Assessment
8. Passive Reconnaissance Results
9. HIPAA/SOC 2 Compliance Gap Analysis
10. 30/60/90-Day Remediation Roadmap
11. Appendix (evidence, payload catalog, methodology)

Deliverables package:

• Assessment Report (PDF, 40+ pages)
• Report (DOCX, editable version)
• Findings Spreadsheet (Excel, filterable)
• Adversarial Payload Catalog (JSON)
• MCP Baseline Snapshots (JSON)
• Scan Artifacts (ZIP)

What We Need From You

We work with whatever you can provide. The more access you grant, the more comprehensive the assessment. Materials can be provided incrementally — we start scanning as each piece arrives.

After the Engagement

Remediation Support

Every finding includes specific, actionable remediation guidance. For critical findings, we provide step-by-step implementation instructions. Our team is available for follow-up questions during the 30 days after report delivery.

Ongoing Monitoring

Many findings represent ongoing risks — MCP servers that could change their tool definitions, new dependencies that could introduce CVEs, log files that accumulate PHI daily. We offer two options for continued coverage:

Self-service scanning — Use your TrustTrace Scan account to run regular scans against your MCP servers and configurations. Baseline comparison (Pro+) catches changes between scans.

Monitoring retainer ($2,500/month) — We run periodic scans on your behalf, monitor for changes, and provide a monthly delta report with new findings and remediation updates.

Reassessment

We recommend a full reassessment every 6–12 months, or after major changes to your AI agent architecture (new agents deployed, new MCP servers connected, framework upgrades).

Assessment clients receive a 20% discount on reassessments within 12 months.

Get Started

Ready to assess your AI agents? Start with a conversation.

→ Contact Us

Or try a free self-service scan first to see what TrustTrace finds before committing to a full engagement.