api-security
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data via
WebFetchandcurlfor the purpose of analyzing API documentation, schemas, and responses. This creates a surface for indirect prompt injection where a malicious API or web server could return content designed to manipulate the agent's behavior.\n - Ingestion points: Retrieval of Swagger/OpenAPI specifications, GraphQL introspection queries, and target API responses as described in
SKILL.mdand the various guide files.\n - Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores instructions embedded within the fetched external data.\n
- Capability inventory: The skill enables high-privilege tools including
Bash(used forcurl,ffuf,kiterunner),Read,Write, andEdit.\n - Sanitization: No sanitization or validation mechanisms are described for the external data before it is interpreted by the agent.\n- [EXTERNAL_DOWNLOADS]: The skill references several external resources and security tools, such as the
HackTrickspentesting wiki and tools likejwt_tool,kiterunner, andffuf. These are recognized as legitimate, well-known resources in the cybersecurity community and are used here for their intended security auditing purposes.\n- [COMMAND_EXECUTION]: The skill includes multiple examples of shell commands and Python scripts designed to perform network-based security audits. These commands are intended to be executed via theBashtool to interact with external target endpoints.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata