OliveTin deliberately gives trusted operators a web interface for running commands. Its safety model therefore depends on preserving the separation between an approved action, the arguments supplied to that action and the output returned from it. CVE-2026-75506 shows how a post-execution hook can break that separation after the main command has already run safely.

The advisory affects versions from 3000.0.0 onward and lists no patched version. The issue is a second-order operating system command injection in shellAfterCompleted. The affected hook places values including command output, ot_username and exitCode into a shell command template processed by sh -c.

The dangerous boundary is after execution

The main action may use safe argument handling or exec mode. That does not protect a later hook if the hook interpolates command output into a shell string. The advisory notes that values in stepExecAfter were not passed through checkShellArgumentSafety, the validation applied to the primary execution path.

The distinction matters. A command can be allowed to emit arbitrary text and still be safe to run. The text becomes dangerous only when a downstream component interprets it as shell syntax. If that downstream interpretation occurs with the OliveTin process's privileges, the original action boundary has expanded into code execution.

The advisory also describes username spoofing through trusted headers as another way to influence ot_username. The header may be accepted because a reverse proxy is expected to set it. The post-execution hook then treats the resulting identity text as shell input, so the trust chain reaches farther than the authentication layer intended.

This is not a complete fix for CVE-2026-27626. The earlier issue addressed safety in the primary execution step. The post-execution step remained on a shell path.

Who is exposed

An attacker needs to influence a value that reaches the post-execution template. In some deployments that may require an authenticated operator who can supply arguments reflected in output. In others, trusted-header spoofing or another output source may provide the injection point. The CVSS score of 6.6 reflects network reach, high complexity and high privileges, not a guarantee that only administrators can exploit every configuration.

Inventory OliveTin actions with shellAfterCompleted or equivalent post-execution commands. Record which template variables they use and whether those variables can contain user input, output from external systems or identity fields derived from a proxy header.

Review the authentication path for ot_username. If a reverse proxy sets that header, ensure clients cannot supply their own copy. Restrict the proxy listener to the expected network path and overwrite the header at the trusted boundary. A header trust review is necessary even if the post-execution hook is later removed.

Risk reduction until a patch

The advisory lists no patched version, so operators need configuration changes in addition to monitoring for an updated release. Disable shellAfterCompleted hooks that interpolate output, usernames or exit codes. Replace them with an exec-mode post-processing action that receives values as separate arguments. If the product does not support an equivalent, perform the post-processing outside OliveTin in a job system with a fixed command and structured inputs.

Do not build a shell string from the output and rely on quoting functions added in the action definition. The quoting must be applied at the final shell boundary or avoided entirely. Different hooks and shells can interpret the same text differently.

Run OliveTin with a dedicated, unprivileged account. Restrict its filesystem and network access to what the approved actions require. This lowers the impact of any injection but does not justify leaving a shell interpolation path active.

Keep trusted-header identity separate from command parameters. Normalize the username, reject control characters and avoid passing it into any command template. If an action needs an identity argument, use a structured field and a fixed argument position.

Detection and response

Search web, proxy and application logs for commands or arguments containing shell metacharacters. The existing task output can make the relevant event hard to find, so include post-execution hook names and action identifiers in logs. Alert on commands that spawn unexpected interpreters, download tools or outbound connections from the OliveTin account.

Review process trees around OliveTin executions. A legitimate action may spawn child processes, so compare against the approved action definition and the expected user. Treat a shell child that only appears after the action completes as a high-priority finding.

If a post-execution injection is confirmed, preserve the hook configuration, task inputs, identity headers and process evidence. Rotate credentials available to the OliveTin account and review hosts reachable from that account. Removing the hook stops the current path, but the injected command may already have persisted elsewhere.

References