Executive brief
Hulumi is a skill pack framework that executes helper scripts during local operations. Versions before 1.3.2 load the threat-model helper script from an unsafe directory path, allowing an attacker to place a malicious file in the workspace that shadows (replaces) the intended helper. An attacker who can modify workspace files can execute arbitrary code during skill execution.
Technical details
The vulnerability is an untrusted search path issue (CWE-426) in which Hulumi resolves the /hulumi-threat-model helper script from an unsafe root directory before checking the intended skill directory. An attacker can place a malicious file with the same name in the workspace directory, causing it to be loaded and executed instead of the legitimate helper during local skill execution. No authentication or special privileges are required; the attack requires only the ability to modify files in the workspace. The vendor patched this in v1.3.2 by anchoring helper resolution to the skill directory, eliminating the shadowing attack surface.
Affected products
- kerberosmansour Hulumi before 1.3.2
Timeline
- 2026-05-15: disclosed: GHSA advisory published
- 2026-05-15: patched: Fix released in v1.3.2
- 2026-08-31: advisory: CVE-2026-82862 published