Application Security · Niklas Schmidt

Penetration Testing & AI Red Teaming

Find vulnerabilities. Validate attack paths. Close the gaps.

I offer authorized penetration testing, security assessments, and AI red teaming for businesses and product teams. My focus is web applications, APIs, and AI agents: which boundaries can be crossed, what data would be affected, and how can the underlying issue be fixed?

Discuss a security assessmentRead security research

Assessment scope

Test security in the system that matters.

01

Application and API penetration testing

Targeted testing of web applications, mobile backends, and SaaS platforms within an agreed scope.

  • Authentication, sessions, roles, and access controls
  • Tenant isolation, business logic, and data access
  • Controlled proofs of concept and vulnerability chaining to validate realistic attack paths
02

AI red teaming and agent security

Adversarial testing of LLM applications, AI agents, and MCP integrations, including the tools and data sources they can access.

  • Direct and indirect prompt injection through documents and tool outputs
  • Data leakage, credential handling, and abuse of tool permissions
  • Assessment of sandbox boundaries and safeguards enforced outside the model
03

Security assessments and remediation

Code and architecture reviews focused on trust boundaries, reachable attack paths, and practical mitigation.

  • Threat modeling and review of security-sensitive data flows
  • Prioritized findings with reproduction steps and technical recommendations
  • Support with patches, regression tests, and verification through retesting

Working together

An agreed scope. Findings you can act on.

  1. 01

    Scope and authorization

    We agree in writing on target systems, test accounts, permitted methods, testing windows, and stop conditions. Testing covers only owned or explicitly authorized systems.

  2. 02

    Assessment and report

    You receive a technical report with findings, impact, reproduction steps, and prioritized recommendations. Critical findings go directly to the agreed point of contact.

  3. 03

    Remediation and retesting

    I can support implementation of the fixes. An agreed retest documents which findings have been resolved and which risks remain open.

Engineering and research

Technical experience from my products and client work.

These projects show security features and architecture decisions I implemented. The research article documents a separate assessment with its methodology, results, and limitations.

Own product

Agent Loadout

Credential vaults, revocable access grants, and a connector that restricts agent access to approved destinations and actions.

View project

Own product

AgentPlan

Private documents, scoped API tokens, and isolated rendering of untrusted HTML content.

View project

Own product

Passcard

Local credential-exposure detection and usage accounting for coding agents without uploading prompts or source code.

View project

Client project · Software development

HeinzerIT Cockpit

Tenant isolation, roles, permissions, audit logs, and AI assistants with scoped data access in a customer portal.

View project

Documented assessment

Ox Alpha: prompt injection and data leakage through reasoning and tool arguments.

A black-box assessment using synthetic test data and unexecuted tool calls. The article documents reproduced model behavior, potential implications for agent systems, and the limits of the findings.

Read security research

What needs testing?

Tell me about your application, its important data, and the security questions you need answered. We can define a suitable scope and the access required for the assessment.

Discuss a security assessment