Advisory record · GHSA-jfpc-wj3m-qw2m

CAI find_file Agent Tool has Command Injection Vulnerability Through Argument Injection

CVSS 9.7 CriticalCriticalEPSS 0.0091 · 60th percentile

Critical, because:

  • CVSS score 9.7: 9.0 or more.

How we decide what’s critical

Published by the GitHub Advisory Database at 30 Jan 2026, 20:38 UTC. agentprime has not analysed this yet, so whether it affects any app is not known.

PackageEcosystemAffectedFixed in
cai-frameworkPyPI<=0.5.10No fixed version yet
Identifiers
GHSA-jfpc-wj3m-qw2m · CVE-2026-25130 · PYSEC-2026-303
Weakness (CWE)
CWE-78
Published
30 Jan 2026
Updated
27 Oct 2026

Machine-written summary

Summary written by Claude Haiku from the source advisory. Not reviewed by a person.

The cai-framework PyPI package, versions up to and including 0.5.10, has a command injection flaw in its find_file agent tool, where user-controlled arguments are passed to a shell via subprocess with shell=True. An attacker could use prompt injection, such as hidden instructions on a webpage the agent reads, to make the agent call find_file with a crafted -exec argument and run arbitrary commands on the host without user approval. The source says a patch exists in the project repository but was not yet published to PyPI, and no fixed PyPI version is listed.

Written 31 Oct 2026.

Not analysed yet

We haven’t published our analysis of this Advisory yet. Once a person has reviewed it, this page will say what’s vulnerable, what has to be true for your app to be affected, and how to check.

Severity and exploitation

ScoreRatingCVSSVectorSource
9.7 Critical 3.1 CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H CNA (GitHub_M)
  • CISA KEV Not listed.
  • EPSS 0.0091, the 60th percentile, on 15 Nov 2026: FIRST’s estimate of the chance of exploitation in the next 30 days.

Timeline

  1. 31 Oct 2026, 14:07 UTC First seen by agentprime
  2. 31 Oct 2026, 14:07 UTC Confirmed by OSV

Sources

Facts on this page come from the GitHub Advisory Database (CC-BY 4.0), OSV and the other sources listed, as published. We don’t reproduce full advisory descriptions, which can contain exploit code.