llamaCVEVulnerability Intelligence
/
Advisory · GitHub · Go · go.probo.inc/probo

GHSA-9FX8-47W4-3VW8

Stored XSS in the console via unsanitized Markdown (iframe srcdoc) in Organization Context; no CSP on the console app

Texto original en inglés · Ver en español

Sin parche publicado
Descargar PNG JSON
CVSS 3.1
4.8
Media
CVSS:3.1/AV:N/AC:L/PR:H/UI:R/S:C/C:L/I:L/A:N Fuente: GitHub
ExplotaciónSin evidenciaNo figura en CISA KEV
Exploits públicos0Ninguno conocido
ParcheSin parcheEl fabricante no ha publicado corrección
Exposición95 dDesde la divulgación, 18 ago 2026, 1:30 a. m.
Detección92 dDesde la divulgación hasta su ingreso al índice
Decisión estimadaSeguir
Remediar en la ventana regular de mantenimiento
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 parcial. No hay versión corregida: aplique mitigaciones.
Sin prisa pero sin pausa, como yo en el páramo.Esta la tienes controlada, confío en ti.
Timeline

Ciclo de vida de la vulnerabilidad

Hoy
Divulgación
Publicada
Explotación
Remediación
Monitoreo
Detectada +1
18 ago 26
18 nov 26
21 nov 26
18 ago 26
11 sept 26
5 oct 26
29 oct 26
21 nov 26
FechaFaseEventoDetalle
Divulgación Publicada GHSA-9FX8-47W4-3VW8 divulgada · CVSS 4.8 (GitHub)
Monitoreo Detectada Ingresó al índice
Monitoreo Verificada Última verificación contra las fuentes
Resumen
Descripción del advisory
Texto original · GitHub · en inglés

Stored XSS in the console via unsanitized Markdown (`<iframe srcdoc>`) in Organization Context; no CSP on the console app

> AI-assistance disclosure: Found and drafted with the assistance of a generative-AI tool
> (Anthropic Claude / "Claude Code"); code references and the payload/sink behavior were reviewed and
> tested by the human reporter against the real, unmodified Probo code before submission.

  • Component: `packages/ui/src/Atoms/Markdown/Markdown.tsx` (the shared Markdown renderer);
  • Version: probod v0.222.2 (`a3b65a644`), default configuration.
  • Class: CWE-79 (Stored Cross-Site Scripting); CWE-693 (missing CSP) as a contributing factor.
  • Severity (proposed): Moderate — CVSS 3.1 `AV:N/AC:L/PR:H/UI:R/S:C/C:L/I:L/A:N` = 4.8
  • Confidence: High on the sink and the absent-CSP gate.

Summary

The shared `Markdown` component renders Markdown with `react-markdown` + `rehype-raw` and no
`rehype-sanitize`, so raw HTML in the content is emitted to the DOM. The Organization Context
field is authored by an OWNER/ADMIN via `updateOrganizationContext` (stored verbatim, no
sanitization) and rendered through this component on a page whose read permission
(`ActionOrganizationContextGet`) extends to VIEWER and AUDITOR. `react-markdown` v10 strips
`on*` handlers and neutralizes `javascript:`/`data:` URLs, but passes an `<iframe srcdoc="…">`
through intact; scripts inside `srcdoc` execute same-origin at the console origin. Because the
console SPA is served without any Content-Security-Policy (verified below), the injected script
executes — a stored XSS that runs in the session of any member (including OWNER and external
AUDITOR) who views the org Context page.

Root cause (source)

Unsafe sink (`packages/ui/src/Atoms/Markdown/Markdown.tsx:29-31`):

<ReactMarkdown
  remarkPlugins={[remarkGfm]}
  rehypePlugins={[rehypeRaw]}   // raw HTML passed through; NO rehype-sanitize
>

`<iframe srcdoc="<script>…</script>">` survives this config (empirically: rendering the payload
through the exact plugin set preserves the `<iframe>` and its `srcDoc` attribute).

Write path: `updateOrganizationContext` (`organization_resolvers.go`) authorizes
`ActionOrganizationContextUpdate` (`core:*` → OWNER/ADMIN) and stores the value verbatim.
Read path: `ContextPage.tsx` renders `<Markdown content={...}/>`; `ActionOrganizationContextGet`
is granted to VIEWER and AUDITOR (`pkg/probo/policies.go:109,134`).

No CSP on the console app (decisive): the CSP `Content-Security-Policy: default-src 'self'` is
set only inside the API server's handler (`pkg/server/api/api.go`, mounted at `/api`). The console
SPA is mounted at `/` (`pkg/server/server.go:80`) and served by `pkg/server/web/web.go` →
`statichandler`, which sets only `Content-Type`, `Cache-Control`, and `ETag`
(`statichandler.go:145-216`) — no `Content-Security-Policy`, no `X-Frame-Options`. The console
`index.html` contains no `<meta http-equiv="Content-Security-Policy">`. So the document that
renders the Markdown has no CSP, and the inline script in the `<iframe srcdoc>` (which inherits the
embedding document's CSP — here, none) executes.

Reachability (hop-by-hop)

  1. Attacker holds OWNER/ADMIN and sets the Organization Context to
  2. A victim member (VIEWER / AUDITOR / another ADMIN / OWNER) opens the Organization Context
  3. No CSP on the console document → the inline script executes same-origin at the console origin

Impact

Stored XSS executing in a victim's authenticated console session. An ADMIN can thereby run script in
the session of an OWNER (theft of the owner's session/CSRF token → act as OWNER — a privilege
escalation independent of the `createUser` bug) or of an AUDITOR (typically an external party
granted limited access; their session may span multiple audited organizations). The absent CSP and
`X-Frame-Options` on the console are a contributing hardening gap (also enabling clickjacking).

Adversarial re-read

  • *"React escapes output."* True for normal JSX, but this sink deliberately opts out via `rehype-raw`
  • *"A CSP blocks the inline script."* Checked and refuted: the `default-src 'self'` CSP is on the
  • *"`react-markdown` strips dangerous HTML."* It strips `on*`/`javascript:` but not `<iframe
  • Residual (reflected in PR:H): the injector must be OWNER/ADMIN. The value is that it crosses to

Preconditions

  • Attacker holds OWNER/ADMIN (to write the context). A victim with read access (VIEWER/AUDITOR/

Remediation

  • Add `rehype-sanitize` after `rehype-raw` in `Markdown.tsx` (or remove `rehype-raw`) — this also
  • Serve a restrictive `Content-Security-Policy` (and `X-Frame-Options: DENY`) on the console SPA
Vector CVSS 3.1

Desglose de métricas

Opción resaltada = valor del vector · izquierda = peor caso · GitHub
Vector de ataqueAV:N
RedAdyacenteLocalFísica
ComplejidadAC:L
BajaAlta
PrivilegiosPR:H
NingunoBajosAltos
InteracciónUI:R
NingunaRequerida
AlcanceS:C
CambiadoSin cambio
ConfidencialidadC:L
AltoBajoNinguno
IntegridadI:L
AltoBajoNinguno
DisponibilidadA:N
AltoBajoNinguno
Score de explotaciónQué tan fácil y probable es que la exploten
6.0/ 10Medio
Vector 6.0 (fácil) · sin evidencia de explotación.
Versiones

Rangos afectados

Afectada Corregida Última
go.probo.inc/probo< 0.257.0
v0.216.1
0.257.0
Remediación

Estado del parche por paquete

Paquete / productoEstadoCorregida enDesde divulgaciónÚltima versión
gogo.probo.inc/probo
Sin parche — — v0.216.1 afectada
Versiones corregidas0.257.0
Explotación

Arsenal público

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

Referencias
Advisoryhttps://github.com/getprobo/probo/security/advisories/GHSA-9fx8-47w4-3vw8

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