latex-thesis-zh
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- COMMAND_EXECUTION (SAFE): The script
scripts/check_format.pyexecutes the external utilitychktexusingsubprocess.run. It correctly passes arguments as a list and avoids the use ofshell=True, which prevents shell injection vulnerabilities even when processing user-defined file paths. - PROMPT_INJECTION (LOW): Detected Category 8 (Indirect Prompt Injection) vulnerability surface. The skill is designed to ingest and process user-controlled content for the purpose of optimization and search.
- Ingestion points: Untrusted data enters the agent context through the parsing of LaTeX (
.tex) and BibTeX (.bib) files inscripts/check_format.py,scripts/verify_bib.py, andscripts/parsers.py. - Boundary markers: Absent. The scripts do not wrap extracted content in delimiters or provide 'ignore embedded instructions' warnings when the data is passed to the agent.
- Capability inventory: The skill possesses the ability to execute subprocesses (
chktex) and generates outputs that suggest further agent tools calls (e.g.,google_web_searchfor missing DOIs inscripts/verify_bib.py). - Sanitization: Absent. There is no logic to sanitize or escape LaTeX comments or text that might contain instructions targeting the LLM's behavior.
Audit Metadata