Results for “tool-poisoning”

Assess AI/ML systems for prompt injection, jailbreak vulnerabilities, model inversion risk, data poisoning exposure, and agent tool abuse, with MITRE ATLAS mapping and guardrail recommendations.

Conduct authorized security assessments of LLM applications and AI agents, covering prompt injection, tool abuse, RAG exposure, memory poisoning, and model supply-chain risks.

Provides a comprehensive penetration testing toolchain with 20+ security tools (Nmap, Nuclei, SQLMap, FFUF, Hashcat, etc.) exposed via MCP servers for authorized vulnerability scanning, exploitation, and reporting.

Probe RAG applications for prompt injection via poisoned retrieved context and embedding manipulation.

Detect and exploit JavaScript prototype pollution vulnerabilities on both client-side and server-side applications to achieve XSS, RCE, and authentication bypass through property injection.

Guides blue-team threat hunting and detection engineering with hypothesis-driven workflows, Sigma/YARA rule creation, SIEM query design, and validation using Atomic Red Team in authorized environments.

Plan and execute authorized vishing (voice phishing) pretext calls to assess employee susceptibility to social engineering and evaluate security awareness controls.