Agent Data Injection Tricks AI Agents into Misreading Tool Data
Agent Data Injection Tricks AI Agents into Misreading Tool Data is a medium-severity prompt injection issue in Agent Data Injection Tricks AI Agents into Misreading Tool D, an AI agent.
Is Agent Data Injection Tricks AI Agents into Misreading Tool D running on your developers' machines? Anomity finds every device with this AI agent installed, shows the version, and can block it by policy. Detect this across your developers' devices with Anomity →
Key facts
- ID
agent-data-injection-tricks-ai-agents-into-misreading-tool-data- Severity
- medium
- Attack type
- Prompt injection
- Component
- Agent Data Injection Tricks AI Agents into Misreading Tool D · AI agent
- Known exploited
- No known exploitation
- Published
- Last verified
What is this vulnerability?
Researchers disclosed Agent Data Injection (ADI), which corrupts the factual data agents trust (sender names, button IDs, tool results) using punctuation-like characters that models misread as real structural markers. In tests it pushed coding assistants into running attacker commands and forged pull-request approval records without hijacking the core task, so any agent that ingests untrusted tool output is exposed.
Am I affected by this vulnerability?
You are affected if Agent Data Injection Tricks AI Agents into Misreading Tool D is installed or configured on a machine. Check the references for the exact versions.
How do I fix it?
Add unguessable random tags to agent-facing field names and treat all tool-output data as untrusted rather than structural.
How do I find every machine running Agent Data Injection Tricks AI Agents into Misreading Tool D?
Anomity inventories the AI agents, MCP servers, IDE extensions and skills on each developer device, so you can see which machines run Agent Data Injection Tricks AI Agents into Misreading Tool D and which version, then enforce a policy on it.