One action from an AI agent rarely looks dangerous on its own. Reading a config file is fine. Calling an outside server is fine. Doing both, one after the other, can be data theft.
That’s the idea behind Project Quarry, a behavior detection engine Cisco has open sourced to help security teams spot rogue AI agents. The Register reports that Cisco wants it to become a “shared language” for agent behavior, much as Snort, which Cisco maintains, did for malicious network traffic and YARA did for malware.
Context over single events
“But what is missing is the entire context under which something may occur,” Amy Chang, Cisco’s head of AI threat and security research, told The Register. She compared it to credit card fraud detection, where a small test charge or a single far-away purchase means little until the pattern emerges.
Quarry looks at the whole agent session. It covers single- and multi-turn LLM conversations, agents, tool use, Model Context Protocol interactions and multimodal inputs. As Cisco’s researchers put it, a policy can produce a finding “when the trace shows a sensitive file read followed within a defined window by an external send.”
What’s in the box
The GitHub repository includes installation instructions, plugins for Claude Code, Codex and other MCP-compatible agents, configuration examples, rule-writing guidance and a starter rule pack. The rule format will look familiar to anyone who has written YARA or Snort rules.
Chang said Quarry would likely have flagged steps in the recent OpenAI-Hugging Face intrusion, such as internet access from a sandbox, or credentials found in a tool result and then used against an outside service. Cisco is clear that rules aren’t a substitute for chain-of-thought monitoring or human oversight.
“We all see slices of the problem, and it’s important for us to share what we see,” Chang said.
Why it matters
As companies hand AI agents real permissions, security teams need ways to monitor what those agents do, and an open standard could speed that up. Expect agent security to show up in more RFPs and compliance checklists.
For security vendors, MSSPs and AI platform providers, agent governance is a fresh category where buyers are still forming opinions. Getting in front of CISOs and AI platform owners now, with practical guidance, is a strong way to build pipeline early.
Sources: The Register, Cisco. Written by the xRapid Media LLP team.
Selling to the IT market? xRapid Media LLP is a High Tech B2B Marketing company. We help technology companies reach decision-makers with verified B2B data, account-based campaigns, email and LinkedIn outreach and appointment setting. Call +91 9767196309 or book a free B2B marketing consultation.