llamaCVEVulnerability Intelligence
/
Advisory · GitHub · Go · github.com/obot-platform/obot

GHSA-X3HF-V55X-JCP6

Access token theft via unvalidated OAuth resource metadata

Parche disponible
Descargar PNG JSON
CVSS 3.1
8.0
Alta
CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:H/I:H/A:N Fuente: GitHub
ExplotaciónSin evidenciaNo figura en CISA KEV
Exploits públicos0Ninguno conocido
Parche1 versiónPrimer parche a -11 d de la divulgación
Exposición21 dDesde la divulgación, 30 oct 2026, 7:36 p. m.
Detección19 dDesde la divulgación hasta su ingreso al índice
Decisión estimadaSeguir de cerca
Monitorear y programar la remediación
Estimación de Llama con el árbol SSVC de CISA a partir de vector CVSS 3.1 y la evidencia de explotación (sin evaluación oficial): sin explotación conocida, no automatizable e impacto técnico total. Hay versiones corregidas disponibles.
No es drama, es prevención. Hazme caso.Mis orejas detectan algo raro por aquí.
Timeline

Ciclo de vida de la vulnerabilidad

Hoy
Divulgación
Publicada
Explotación
Remediación
Parche v0.27.0
Monitoreo
Detectada +1
20 oct 26
31 oct 26
19 nov 26
21 nov 26
20 oct 26
28 oct 26
5 nov 26
13 nov 26
21 nov 26
FechaFaseEventoDetalle
Remediación Parche v0.27.0 github.com/obot-platform/obot · -11 d desde divulgación
Divulgación Publicada GHSA-X3HF-V55X-JCP6 divulgada · CVSS 8 (GitHub)
Monitoreo Detectada Ingresó al índice
Monitoreo Verificada Última verificación contra las fuentes
Resumen
Descripción del advisory
Texto original · GitHub · en inglés

Summary

When Obot connected to a remote MCP server that requires OAuth, it trusted the `resource` identifier in the server's protected resource metadata without checking that it matched the server. A malicious or compromised remote MCP server could name a different service as the resource, point Obot at an authorization server the victim already uses, and have Obot obtain an access token for that other service. Obot then sent the token to the malicious server with each request, giving the attacker a working token for a service they never touched.

Am I affected?

You are affected if you run Obot >= v0.26.0, <= v0.26.3 and users connect to remote MCP servers that use OAuth. The attack also requires all of the following:

  • The remote MCP server is malicious or has been compromised.
  • The authorization server the attacker names accepts Obot as a client, through either open client registration or client ID metadata documents, and honors RFC 8707 resource indicators. Many enterprise identity providers support neither, and Obot doesn't send resource indicators to Microsoft Entra ID.
  • The victim approves (or has previously approved) the consent prompt from that authorization server.

Details

RFC 9728 requires an OAuth client to reject protected resource metadata whose `resource` value doesn't match the resource it requested. Obot only checked that the field was present, then used the value as the resource indicator in the authorization request and stored the resulting token against the connection to the remote server. The authorization server and requested scopes also came from the remote server, so the attacker chose which authorization server issued the token, which service it was for, and what permissions it requested.

Impact

An attacker could obtain a valid access token, issued to the victim, for another service protected by the same authorization server. The attacker chooses the requested scopes, so the token could grant the full read and write access the victim is able to consent to on that service. The victim sees only a legitimate consent screen from an identity provider they trust.

Mitigation

Upgrade to v0.27.0 or later. Obot now rejects protected resource metadata unless:

  • the resource has the same origin as the MCP server URL, and its path is the server's path or a parent of it,
  • any query on the resource matches the server URL's query exactly,
  • the resource has no fragment, and
  • neither URL's path contains anything that servers might normalize differently: a `.` or `..` segment, percent-encoding, `\`, or `;`.

Severity

CVSS v3.1 Score: 8.0/10 (High) — `CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:H/I:H/A:N`

Credits

The Obot team would like to thank @Srinivasan-Jayaraman https://github.com/Srinivasan-Jayaraman for responsibly disclosing this issue in accordance with our security policy https://github.com/obot-platform/obot/?tab=security-ov-file.

Vector CVSS 3.1

Desglose de métricas

Opción resaltada = valor del vector · izquierda = peor caso · GitHub
Vector de ataqueAV:N
RedAdyacenteLocalFísica
ComplejidadAC:H
BajaAlta
PrivilegiosPR:N
NingunoBajosAltos
InteracciónUI:R
NingunaRequerida
AlcanceS:C
CambiadoSin cambio
ConfidencialidadC:H
AltoBajoNinguno
IntegridadI:H
AltoBajoNinguno
DisponibilidadA:N
AltoBajoNinguno
Score de explotaciónQué tan fácil y probable es que la exploten
5.9/ 10Medio
Vector 5.9 (moderada) · sin evidencia de explotación.
Versiones

Rangos afectados

Afectada Corregida Última
github.com/obot-platform/obot≥ v0.26.0 ≤ v0.26.3
v0.26.0
v0.26.3
v0.28.1
Remediación

Estado del parche por paquete

Paquete / productoEstadoCorregida enDesde divulgaciónÚltima versión
gogithub.com/obot-platform/obot
Corregida v0.27.0 -11 d v0.28.1 no afectada
Versiones corregidasv0.27.0
Explotación

Arsenal público

Sin exploits, módulos ni PoC públicos conocidos a la fecha de verificación.

Referencias
Advisoryhttps://github.com/obot-platform/obot/security/advisories/GHSA-x3hf-v55x-jcp6

Las referencias externas se muestran para copiar: cve.llama.ec no lo redirige a otros sitios.