# CVE-2026-37003: Agno prompt injection remote code execution in PythonTools and ShellTools

Source: https://junglewise.ai/threats/cve-2026-37003-agno-prompt-injection-remote-code-execution-in-pythontools-and  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-08-27. Updated: 2026-09-25  
License: CC BY 4.0 with a link back

CVE-2026-37003 · Severity: critical · CVSS 9.8 · Published 2026-08-27

Agno is an AI agent platform that allows organizations to build autonomous agents for customer support, sales, research, and internal workflows. Versions up to 2.5.8 contain a critical vulnerability where malicious instructions embedded in web pages or documents processed by the agent can trigger arbitrary code execution on the host server, potentially compromising the entire agent platform and any data it accesses.

Technologies: [Agno](https://junglewise.ai/threats/technologies/agno). Vendors: [Agno](https://junglewise.ai/threats/vendors/agno).

## Quick answers

**Is CVE-2026-37003 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-37003 being exploited in the wild.

**How severe is CVE-2026-37003?** Critical severity, with a CVSS score of 9.8, published 27 August 2026.

**Which products and versions are affected?** Agno up to and including 2.5.8.

**Is there a fix for CVE-2026-37003?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-37003 attacked?** It is attacked over the network.

**Am I affected by CVE-2026-37003?** Check whether you run Agno at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

The vulnerability is a prompt injection attack affecting Agno's PythonTools and ShellTools components. These tools accept LLM-generated arguments without sanitization and pass them directly to dangerous execution sinks including Python's exec(), runpy.run_path(), and subprocess.run(). An unauthenticated attacker can exploit this by injecting malicious instructions into content (such as web pages or documents) that the Agno agent processes, achieving arbitrary Python code execution and OS command execution on the host server. No authentication is required, and the attack is triggered through normal agent operation. Patches for this critical vulnerability should be available or forthcoming.

## Affected products

- Agno Agno up to and including 2.5.8

## Timeline

- 2026-08-27: disclosed: Published on NVD

## References

- [http://agno.com/](http://agno.com/)
- [https://yerangamage.com/cves/detail/?slug=agno-rce-tools](https://yerangamage.com/cves/detail/?slug=agno-rce-tools)

## Related threats

- [Agno unsafe reflection in PandasTools](https://junglewise.ai/threats/cve-2026-119050-agno-through-3-1-11-contains-an-unsafe-reflection-vulnerability)
- [Agno SQL injection in ClickHouse vector database backend](https://junglewise.ai/threats/cve-2026-10105-agno-sql-injection-in-clickhouse-vector-database-backend)
- [Agno eval injection in model execution component](https://junglewise.ai/threats/cve-2026-35002-agno-eval-injection-in-model-execution-component)