Published · Preprint

Indirect Prompt Injection in LLM Agent Frameworks: A Comparative Study

A controlled empirical study across LangChain, MCP, and the Anthropic API. 3,274 trials measuring how injected instructions propagate through multi-step agent pipelines. Task framing, not framework architecture, drives vulnerability.

Across 3,274 controlled trials, task framing was found to drive vulnerability more than framework architecture. Instruction-following tasks reached 91% injection success rates under adversarial framing, with success correlating to instruction authority framing rather than payload complexity.

Hardened system prompts achieved 100% IPI mitigation across all tested frameworks. MCP adds no measurable attack surface relative to the raw API under equivalent task structures. Dataset and framework are open-sourced on GitHub and Zenodo.

Preprint submitted · arXiv link forthcoming
In Progress
Active
Turn depth and secure tool call degradation in agentic pipelines
Next
AI agent CVE landscape paper. Patterns, root causes, and disclosure timelines across the ecosystem.
Frameworks Studied
LangChain
MCP (Model Context Protocol)
Anthropic API
CrewAI
AutoGen