api-test-suite-builder
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for scanning source files using shell commands to detect API routes. This creates a surface where a maliciously crafted repository could use specific file names or contents to influence the generated test output or cause errors in the extraction process.\n
- Ingestion points: File paths and route definitions scanned via
findandgrepcommands inSKILL.md.\n - Boundary markers: None; the logic does not include delimiters or warnings to ignore adversarial content within the codebase being scanned.\n
- Capability inventory: The skill utilizes subprocess calls including
find,grep,sed, andxargs.\n - Sanitization: Sanitization relies on basic regex patterns (
grep -oE), which may not account for complex adversarial file system structures.\n- [COMMAND_EXECUTION]: The skill provides multiple shell snippets for route detection and testing. While these are common developer tools, executing these commands on untrusted source code involves risks associated with shell interpolation and command processing.\n- [CREDENTIALS_UNSAFE]: The Python test example inEXTENDED.mdcontains a hardcoded credential:JWT_SECRET = \"test-secret\". Although this is intended for a testing environment and explicitly labeled as such, hardcoding secrets in source code files is a security anti-pattern.
Audit Metadata