The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external input through its automation scripts, creating a surface for indirect prompt injection.\n
Ingestion points: scripts/api_scaffolder.py parses OpenAPI specifications (YAML/JSON), and scripts/database_migration_tool.py parses SQL schema files provided by the user.\n
Boundary markers: The scripts do not implement delimiters or instructions to ignore or sanitize embedded instructions within the processed data.\n
Capability inventory: api_scaffolder.py generates and writes TypeScript route handlers and validators to the local filesystem; database_migration_tool.py generates and writes SQL migration scripts.\n
Sanitization: Data from external specifications is interpolated into code templates with minimal sanitization, potentially allowing an attacker to inject malicious code or logic into the generated output files.\n- [COMMAND_EXECUTION]: The skill includes utility scripts that perform network operations and file system modifications.\n
scripts/api_load_tester.py executes network requests to user-provided URLs. This could be leveraged to perform internal network scanning or Server-Side Request Forgery (SSRF).\n
scripts/api_load_tester.py contains a --no-verify-ssl flag that explicitly disables SSL certificate verification (ssl.CERT_NONE), exposing the agent to man-in-the-middle attacks if used against untrusted or production endpoints.

senior-backend