open-banking-attack

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill interacts with external Open Banking APIs to fetch configurations and account data, creating a surface for indirect prompt injection.\n- Ingestion points: Data is retrieved from endpoints like .well-known/openid-configuration and /accounts in SKILL.md and guides/open-banking-attack-playbook.md.\n- Boundary markers: No specific delimiters or warnings are provided to the agent to treat this external data as untrusted.\n- Capability inventory: The skill uses curl, httpx, and python3 for network and file operations.\n- Sanitization: There is no explicit sanitization of API responses before processing.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download security tools from well-known public repositories, such as the OIDF conformance suite and UK OBIE sandbox components.\n- [COMMAND_EXECUTION]: Extensive use of shell commands via curl, openssl, and jwt-tool is documented throughout payloads.md and SKILL.md for reconnaissance and exploitation purposes.\n- [REMOTE_CODE_EXECUTION]: The playbook in guides/open-banking-attack-playbook.md includes procedures for cloning, building, and executing code from external repositories (e.g., mitreid-connect, OpenBankingUK) to set up a security testing environment.
Recommendations
  • Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:36 PM