CVE-2026-82447
Skyvern before 1.0.45 contains a sandbox escape vulnerability in TextPromptBlock that renders prompts twice, first through a sandboxed Jinja environment and then through an unsandboxed environment. Attackers can inject malicious Jinja template syntax through workflow parameters or upstream block output to execute arbitrary code with server process privileges.
Scoring
- Severity
- HIGH
- CVSS base score
- 8.8
- CVSS vector
- CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N
- EPSS probability
- 0.45%
- CWE
- CWE-1336
- Published
- 2026-08-29
- Last modified
- 2026-09-01
Affected products
- Skyvern-AI skyvern
Weakness type
Related vulnerabilities
- CVE-2026-19584 — Velociraptor VQL injection during notebook restore from backup
- CVE-2026-87021 — Tanium addressed an unauthorized code execution vulnerability in Comply.
- CVE-2026-33387 — Insufficient sanitization of Dashboards in Guardian/CMC before 26.3.0
- CVE-2026-75650 — Adobe Commerce | Improper Neutralization of Special Elements Used in a Template Engine (CWE-1336)
- CVE-2026-52762 — YesWiki: Authenticated (Admin) Server-Side Template Injection to Remote Code Execution via Bazar Semantic Templates
- CVE-2026-46636 — Twig: Sandbox method allowlist bypass via `Markup` subclass
- CVE-2026-85654 — Code injection in the CDK generator in Amazon awslabs.dynamodb-mcp-server
- CVE-2026-13297 — Security vulnerabilities have been addressed in IBM Verify Identity Access and IBM Security Verify Access