OpenOps CVE-2025-68922

Requires a valid, authenticated low-privileged OpenOps account able to create or edit a workflow that uses the Terraform block, then supply attacker-controlled HCL content, file paths, or resource/property identifiers that are passed server-side to the hcledit CLI; the injection occurs because filePath and resource type/name/property values were interpolated into shell commands without escaping (fixed in 0.6.11). Limiting factors: authentication is required (PR:L), the Terraform block feature must actually be invoked, and per the CVSS vector exploitation involves user interaction (UI:R) and high attack complexity (AC:H) because malicious identifiers must survive hcledit parsing/re-emission. Only versions before 0.6.11 are affected.

CVSS 7.4 (High) reflects a server-side OS command injection (CWE-78) in the OpenOps Terraform block, confirmed by the vendor patch (PR #1767) which adds shell-argument escaping (escapeShellArgument) and strict identifier validation around hcledit CLI invocations - so the root cause is well established. However, exploitation signals are modest: the NVD vector is AV:L/AC:H/PR:L/UI:R, meaning it requires an authenticated low-privileged account, a non-trivial crafting effort, and user interaction, which tempers the raw score. No EPSS probability, no CISA KEV listing, and no public proof-of-concept were provided, so this is 'no public exploit identified at time of analysis' rather than active exploitation. A notable conflict: the vector lists AV:L (local) even though OpenOps is a network-accessible workflow-automation platform where a Terraform block is configured and run server-side - AV:L likely reflects the attacker supplying local-style template/file content, but the practical reach may be network-adjacent; verify the intended vector with the vendor. Net: a legitimate patch-worthy RCE for OpenOps operators, but a prioritized-but-not-emergency issue absent any exploitation evidence, bounded by authentication and the AC:H/UI:R constraints.

An authenticated OpenOps user with permission to build workflows creates a Terraform block containing a crafted HCL template or file path whose resource/property identifiers embed shell metacharacters. When OpenOps processes the block through the hcledit CLI, the unescaped values are interpolated into a shell command and the injected OS commands execute on the server host. No public exploit exists, but the fix diff clearly reveals the unescaped filePath and resource-identifier injection points.