Capgo CVE-2026-56230

Requires a valid authenticated Capgo account (PR:L) and knowledge of another tenant's limited-key identifier, which the attacker submits in the client-controlled x-limited-key-id HTTP header to any endpoint protected by middlewareKey(). The exploit works against default configurations of affected versions (before 12.128.2) with no special feature flag needed, as long as the limited-key mechanism is in use. Limiting factors: the attacker must already authenticate to the API and must obtain or enumerate a valid target key ID, and impact is confined to the Capgo platform's own resources (SC:N) rather than enabling lateral movement into other systems.

The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H, base 8.7) indicates a network-reachable, low-complexity attack that needs only a low-privilege authenticated account and no user interaction, with high confidentiality, integrity and availability impact on the vulnerable system - a credible high priority for any multi-tenant deployment. There is no EPSS score, no CISA KEV listing, and no published proof-of-concept in the provided data, so active exploitation is not confirmed; the realistic risk is gated by the need for a valid authenticated account and knowledge of a target tenant's limited-key ID. Subsequent-system impact is None (SC:N/SI:N/SA:N), meaning blast radius is data and actions within the Capgo platform rather than a pivot into other systems. Net assessment: a genuine, easily-triggered tenant-isolation break that warrants prompt patching, but absent KEV/POC/EPSS signals it reads as a vendor-disclosed (VulnCheck/GHSA) issue rather than one under demonstrated mass exploitation.

An attacker registers or already holds a low-privilege Capgo account, learns or guesses the limited-key identifier of a victim tenant, and issues normal API requests with the x-limited-key-id header set to that victim key. Because middlewareKey() accepts the header without an ownership check, the requests execute against the victim tenant's resources, letting the attacker read, modify, or push updates across the tenant boundary. No public proof-of-concept is referenced, but the low attack complexity (AC:L) and lack of required user interaction make reliable exploitation straightforward once a target key ID is obtained.