The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text to customize LaTeX templates and inspects external PDF files, which constitutes a surface for indirect prompt injection.
Ingestion points: scripts/customize_template.py ingests user-provided arguments for --title, --authors, and --affiliations. scripts/validate_format.py ingests PDF files for inspection.
Boundary markers: No specific delimiters or warnings for embedded instructions are present in the processing scripts.
Capability inventory: scripts/customize_template.py has file-writing capabilities (open(path, 'w')). scripts/validate_format.py has command execution capabilities via subprocess.run to call PDF inspection tools.
Sanitization: The customize_template.py script performs direct regex substitution of user strings into LaTeX files without explicit sanitization for LaTeX special characters or commands.
[COMMAND_EXECUTION]: The script scripts/validate_format.py executes external system commands using the subprocess module.
Evidence: The script calls pdfinfo and pdffonts (part of the Poppler utility suite) to extract metadata and font information from PDF files.
Risk Mitigation: The implementation uses list-based arguments in subprocess.run() and does not use shell=True, which effectively mitigates common shell injection vectors.

venue-templates