Svelte devalue EUVD-2026-119453

Exploitation requires the application to pass untrusted, attacker-influenced data to devalue's uneval function AND to subsequently eval the generated code string - this specific serialize-then-eval usage pattern of devalue <= 5.6.2 is the core precondition. Limiting factors: the advisory states the flaw only manifests 'under certain circumstances' (reflected as AC:H/AT:P in the CVSS 4.0 vector), impact is confined to Low confidentiality/integrity (polluted prototypes changing the reconstructed data's shape, CWE-1321) with no availability impact, and there is no attacker privilege or user interaction requirement beyond supplying the untrusted data. Applications that never eval devalue's uneval output, or that never feed untrusted data into uneval, are not exploitable.

This is a genuinely low-priority issue, and the signals are internally consistent rather than conflicting. The vendor-supplied CVSS 4.0 vector scores only 2.1 with AC:H (exploitation only occurs 'under certain circumstances') and AT:P (a specific attack requirement must be met), and caps all impacts at Low (VC:L/VI:L/VA:N, SC:L/SI:L/SA:N) - consistent with a prototype-pollution integrity/shape issue rather than code execution. The vuln.today decision label is 'monitor', the lowest urgency tier, which agrees with the low score. No EPSS score, no CISA KEV listing, and no proof-of-concept exploit were provided in the input; the only references are the GHSA advisory and the VulnCheck advisory, which are disclosure writeups, not working exploit code. Exploitation status: no public exploit identified at time of analysis. The practical ceiling of this flaw is that eval'd output objects acquire polluted prototypes so the reconstructed data differs in shape from the input (CWE-1321); whether that escalates to anything more serious depends entirely on how the consuming application uses the deserialized object, and the available sources do not establish any such escalation. One caveat on the input data: the supplied vector uses PR:H, but the described trigger is an attacker supplying untrusted data to an application that chose to uneval/eval it - that is an application-design/deployment precondition, not an attacker privilege, so the realistic attacker privilege is closer to PR:N gated by AC:H/AT:P; this does not change the Low severity conclusion.

An application uses devalue's uneval to serialize data that includes attacker-controlled input and later eval's the generated code string to reconstruct it. The attacker supplies crafted untrusted data whose serialized form, once eval'd, creates objects with polluted prototypes, so the rehydrated object has a different shape than intended. No proof-of-concept exploit was provided, and the attack only works in the specific circumstance where untrusted data flows into uneval and the output is subsequently eval'd.