prd
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it interpolates untrusted user input (feature descriptions) directly into generated files without using delimiters or sanitization. Malicious instructions could be embedded in the output document to influence other agents or developers who later read the PRD.\n
- Ingestion points: User-provided feature descriptions used to generate the PRD (SKILL.md).\n
- Boundary markers: None; the instructions do not specify any markers (like quotes or blocks) to isolate user content from the template.\n
- Capability inventory: Filesystem write operations are used to save output to the 'tasks/' directory.\n
- Sanitization: No input validation or character escaping is present.\n- [COMMAND_EXECUTION]: The skill performs filesystem write operations to create markdown files in the 'tasks/' directory. The filenames are derived directly from user input (feature-name), which could potentially be used for path traversal or file overwriting if the user provides a manipulated name.
Audit Metadata