Terragrunt CVE-2026-45099

Exploitation requires that the victim's Terragrunt configuration references a remote module source that is either directly controlled by the attacker or has been compromised to serve a forged .terragrunt-module-manifest. The vulnerability triggers automatically during normal Terragrunt initialization - no additional user interaction is required once the Terragrunt command is invoked. File deletion is bounded by the OS-level permissions of the Terragrunt process; it cannot delete files the process user lacks access to, and it cannot escalate privileges. Exploitation is not possible when only local modules are used, or when all remote module sources are pinned to verified, immutable Git SHAs from internally controlled repositories. The flaw does not affect confidentiality (no read or exfiltration capability) and does not provide code execution. Non-default usage (consuming external, potentially untrusted module sources) is required - organizations using Terragrunt exclusively with internal, SHA-pinned modules are not exposed.

No CVSS score or EPSS probability is available from NVD or the vendor advisory, requiring defenders to perform independent assessment. The attack vector is network-delivered (a remote module must be fetched), with no required attacker privileges on the victim system, but exploitation is gated on user interaction: the victim must explicitly run a Terragrunt command that triggers module download against an attacker-controlled source. This narrows real-world risk significantly - the attack surface is limited to organizations or individuals who consume remote modules from external, potentially untrusted sources, which is not an uncommon pattern in public registry usage. The impact is limited to file deletion (no RCE, no exfiltration, per vendor confirmation), but in CI/CD pipelines this can translate to high availability and integrity impact - deletion of deployment keys, pipeline scripts, or source code. The absence of a KEV listing and lack of any known public exploit suggest exploitation has not been observed in the wild at time of analysis. The 'RCE' tag in the source metadata is inconsistent with the vendor advisory and should be treated as a mis-classification. Overall risk is meaningful in automated environments consuming third-party modules but is substantially mitigated by the prerequisite of module source trust.

An attacker publishes a malicious Terraform module to a public registry - or compromises an existing trusted module - and embeds a crafted .terragrunt-module-manifest containing traversal paths such as ../../.ssh/authorized_keys or ../../app/deploy.sh. When a CI/CD runner or developer executes any Terragrunt command that triggers module initialization against this source, Terragrunt automatically downloads the module, reads the forged manifest during cleanup, and silently deletes the targeted out-of-scope files on the host filesystem without any further user interaction. In a pipeline environment, this could remove deployment credentials, pipeline configuration files, or source code artifacts, halting the deployment workflow and potentially requiring manual recovery.