ash_typescript EUVD-2026-97336

Exploitation requires the attacker to reach the ash_typescript RPC layer (HTTP or channel) and provoke an error for which the error builder has no dedicated clause - concretely, either (1) passing a struct-typed action argument with an invalid nested value, causing cast_map_to_struct/2 in input_formatter.ex to throw a raw Ash.Type cast error into the builder, or (2) performing field selection on a generic action that returns a primitive, causing format_field_type/1 to inspect the full Ash.Info.Manifest.Type struct into vars.returnType. No authentication is required (PR:N) and no user interaction (UI:N), but the CVSS 4.0 AT:P flag reflects a genuine limiting factor: the attacker must deliberately hit these specific unhandled pipeline-error paths rather than any generic error, since Ash exceptions take a separate, redacted path. Exploitation is further limited to applications actually exposing ash_typescript RPC actions, and disclosure is confined to internal Elixir terms - resource module names (vars.resource) are explicitly out of scope and not considered sensitive.

This is a medium-severity information disclosure flaw (CWE-209), and the available signals are internally consistent rather than conflicting. The vendor CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:N/UI:N) confirms network-reachable, unauthenticated exploitation but with an Attack Requirement present (AT:P) - meaning the attacker must specifically provoke an error path the builder has no clause for, not trigger it trivially against any request. Impact is bounded: only VC:L (low confidentiality), with no integrity or availability impact and no scope change, so the realistic outcome is leakage of internal Elixir terms (cast errors, Ash.Info.Manifest.Type structs) in RPC error responses, not code execution or data modification. No EPSS score, no CISA KEV listing, and no public proof-of-concept are present in the provided data, so there is no public exploit identified at time of analysis. The description explicitly scopes OUT resource module names (vars.resource) as non-sensitive, which narrows the blast radius further. Net assessment: a legitimate issue worth patching for teams running affected versions, but not an urgent, mass-exploited priority - it is a disclosure-only, bounded-impact bug in a niche Elixir library, consistent with the 6.3 medium rating rather than an inflated high-CVSS/low-real-risk case.

An unauthenticated attacker sends a crafted RPC request to an application built on ash_typescript - for example supplying a struct-typed action argument with an invalid nested value, or requesting field selection on a generic action that returns a primitive - to deliberately trigger an error path that AshTypescript.Rpc.ErrorBuilder.build_error_response/1 has no dedicated clause for. The builder then inspect/1's the raw error term (a cast error or the full Ash.Info.Manifest.Type struct) into details.error or vars.returnType without the redaction applied to Ash exceptions, leaking internal Elixir terms back in the HTTP or channel response. No public exploit code is identified at time of analysis, and the CVSS 4.0 AT:P requirement means the attacker must specifically provoke the unhandled case.