format-resume
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local script (
format_cv.py) usingsubprocess.runto perform document formatting. This execution is used for its intended purpose of document generation and uses argument lists to avoid shell injection risks. - [PROMPT_INJECTION]: The skill processes untrusted resume/CV content provided by the user, which presents an indirect prompt injection surface.
- Ingestion points: User-pasted raw text or local file content provided at runtime (SKILL.md, Step 1).
- Boundary markers: The skill does not explicitly use delimiters or include 'ignore instructions' warnings when the agent processes the untrusted resume content.
- Capability inventory: The skill can write to the local file system (JSON, DOCX, PDF, and YAML configuration files) and execute local Python scripts via subprocess calls (SKILL.md, Steps 5, 7, and 8).
- Sanitization: No explicit sanitization of user input is described before it is parsed for semantic structure.
- Mitigation: A user confirmation step (SKILL.md, Step 3) is built into the workflow, requiring the user to verify the agent's interpretation of the resume data before any formatting or script execution proceeds, significantly reducing the risk of automated exploitation.
Audit Metadata