Zero Radar
← feed
CriticalVulnerabilityCVE-2026-119050updated 29 Oct 2026, 18:18 UTC

Agno AI agent tool can be tricked into running code or reading local files

Agno is a framework for building AI agents. A flaw lets someone hide instructions in text the AI reads, tricking the agent into running dangerous data-loading functions that can execute code on the server or read files from it.

Official, 2 sources
SeverityCriticalCVSS 9.2 of 10
Being exploited?UnclearNo reports either way yet
Fix available?UnclearNo fixed version published in the sources yet
How sure is this?Official, 2 sourcesBacked by an official source

How the attack works simplified

  1. 1Attacker plants promptHidden instructions placed in content the agent reads
  2. 2Agent calls PandasToolsModel supplies a function name to the tool
  3. 3getattr invokes pandasUnsafe reflection runs any pandas function
  4. 4read_pickle or readers runLoads attacker-controlled paths or local files
  5. ResultAttacker runs code or reads local files

What happened

In Agno through 3.1.11, the PandasTools component passes model-controlled function names to getattr(), allowing invocation of arbitrary pandas functions such as read_pickle and other file readers.

Why it is dangerous: An attacker using prompt injection can achieve code execution on the host running the agent, or read local files. Anyone running an affected Agno agent that processes untrusted input is at risk.

Are you affected?

You are affected if you run Agno through version 3.1.11 with PandasTools enabled and the agent processes untrusted input.

Agno through 3.1.11 (PandasTools component)

What to do

  1. Check whether your Agno version is 3.1.11 or earlier and whether PandasTools is enabled
  2. Disable PandasTools or restrict agent access to untrusted input until a fixed version is confirmed
  3. Update Agno to the fixed version listed in the GitHub advisory GHSA-jq64-vhh4-4hvv once available

Timeline

  1. 29 Oct 2026, 14:17 UTC
    Publicly disclosed
  2. 29 Oct 2026, 14:17 UTC
    Reported by NVD · NVD
  3. 29 Oct 2026, 14:19 UTC
    Picked up by Zero Radar
  4. 29 Oct 2026, 18:18 UTC
    GitHub Advisory reported on it · GitHub Advisory

Sources 2 reports, oldest first

Official vendor, government CERT or advisory database

Technical details

ID
CVE-2026-119050 · GHSA-jq64-vhh4-4hvv
CVSS
9.2 / 10
EPSS
0.4% chance of exploitation within 30 days
Disclosed
29 Oct 2026, 14:17 UTC
On Zero Radar
29 Oct 2026, 14:19 UTC (2 min after disclosure)
Last change
29 Oct 2026, 18:18 UTC

The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works

Agno AI agent tool can be tricked into running code or reading local files · Zero Radar