Activepieces CVE-2026-73081

Exploitation requires an authenticated session on the Activepieces platform with permission to create or edit flows (CVSS 4.0 PR:L confirmed). The trigger is the worker's code-compilation step, which fires when a Code step is processed; placing shell metacharacters anywhere in the step's name field is sufficient to inject commands. No special server configuration, non-default feature flag, or specific platform version beyond 'prior to 0.80.0' is required. Exploitation is bounded by the flow creation/editing permission: users without that role cannot submit step names to the compilation pipeline. The code sandbox - Activepieces' intended isolation mechanism for Code steps - does not limit this attack because the injection executes during compilation, before the sandbox is instantiated. Post-exploitation reach is further bounded by what internal network segments and services the worker process has access to.

The CVSS 4.0 score of 8.7 with VC:H/VI:H/VA:H reflects full compromise of the vulnerable system (worker process) with read, write, and availability impact. The subsequent-system metrics SC:L/SI:L/SA:L confirm limited but real lateral reach, consistent with the description's reference to internal services accessible from the worker. The most significant risk amplifier is PR:L: any authenticated user - not just administrators - who can create or edit flows can exploit this, dramatically broadening the attacker surface in multi-user or SaaS deployments. AC:L and AT:N indicate no special timing, race, or configuration prerequisites. The pre-sandbox execution context means that even hardened sandbox configurations provide no defense. There is no CISA KEV listing and no confirmed POC at time of analysis, which reduces immediate urgency slightly, but the trivial exploitation mechanics (inject metacharacters into a name field) and severe outcome (secret exfiltration, internal network access) make this a genuine remediation priority for any deployment where untrusted or semi-trusted users hold flow-editing permissions. No conflicting signals between CVSS, description, and CWE were identified.

An authenticated Activepieces user with flow-editing access creates or modifies a Code step and sets its name to a value containing shell metacharacters, such as a semicolon followed by a curl command that exfiltrates the contents of the worker's environment variables to an attacker-controlled server. When the platform compiles the flow, the worker process interpolates the malicious step name into its build command, executing the injected payload before any sandbox is applied. The attacker receives the worker's environment secrets - including API keys, database credentials, or internal service tokens - and can chain further commands to write backdoors or probe internal network services.