skills/baberarjumand/technical-resume-generator_agent-skill/tech-resume-generator/Gen Agent Trust Hub
tech-resume-generator
Warn
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documents from the user_data/ directory to extract career history and metrics. Ingestion points: The scripts/extract_user_data.py script reads various file types (PDF, MD, TXT, Images) and passes the extracted content to the agent context. Boundary markers: Absent; there are no instructions or delimiters to separate untrusted file content from the skill's instructions. Capability inventory: The agent has the ability to execute shell commands and write files to the workspace. Sanitization: Absent; the content is processed and compiled into professional_data.md without filtering for potential prompt injection.
- [DYNAMIC_EXECUTION]: The SKILL.md file explicitly instructs the agent to handle 'Unknown formats' by writing new Python scripts to scripts/extractors/ and then executing them. This runtime code generation and execution on user-provided data presents a security risk, as a malicious user could provide a file designed to influence the agent into generating and running harmful code.
- [COMMAND_EXECUTION]: The scripts/extract_user_data.py and scripts/download_sample_resumes.py scripts utilize the subprocess module to execute system-level commands including pdftotext, pdfinfo, pdftoppm, and curl. While used for the intended processing and asset acquisition functionality, this represents a sensitive capability that operates on user-provided file paths.
- [EXTERNAL_DOWNLOADS]: The scripts/download_sample_resumes.py script fetches resume assets from various external domains such as university websites (UNC, Yale, Harvard), GitHub, and Google Docs. These operations are conducted via curl and are intended for acquiring research sample data.
Audit Metadata