SecEng Workbench
SecEng Attack
Attack.
Run adversarial tests against prompts, agents, tools, retrieval, policies, and model behavior.
SecEng Attack is the adversarial evaluation harness for AI systems. Test against the attack paths that matter — prompt injection, jailbreaks, tool abuse, data leakage, RAG poisoning, and agent authority abuse. Map every result to OWASP LLM, MITRE ATLAS, and NIST AI RMF.
Capabilities
What Attack instruments do.
Direct & indirect prompt injection
Test user-input injection, document injection via RAG, email and ticket injection, and template manipulation. Reproduce the full injection surface with evidence and replayable traces.
Agentic tool abuse
Simulate delegated authority abuse, unsafe tool chains, approval bypass framing, and unauthorized external actions across agent workflows and MCP server integrations.
RAG corpus poisoning
Seed poisoned documents, hostile chunks, and indirect prompt injection into the retrieval corpus. Confirm whether hostile content influences model responses before it reaches production.
Jailbreaks and policy bypass
Test instruction hierarchy violations, role-play bypasses, token manipulation, and structured adversarial framing against system prompt protections and policy guardrails.
Multimodal and synthetic media
Exercise OCR, EXIF, steganography, image prompt injection, and synthetic-media abuse paths. Test output safety where users actually interact with the model.
Regression and coverage reporting
Generate replayable test fixtures from every confirmed exploit. Track scenario coverage, gap analysis, and ATLAS/NIST rollups for every AI feature release.
Instruments
Attack instruments.
SecEng Adversarial Range
157 adversarial scenarios across 15 attack packs and 22 threat vectors. Fixture-driven, evidence-grade, framework-mapped.
SecEng Artifact Analyzer
Triage Rust, Go, browser, and MCP artifacts into artifact authority evidence, risk signals, and analyst next steps.
SecEng RAG Test Harness
Retrieval authorization, corpus poisoning, tenant boundary, and leakage tests for RAG systems.
SecEng Authority Graph
Build agent authority graphs. Detect dangerous tool compositions and approval bypass paths before they reach production.
Standards alignment