The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/validate_format.py uses the subprocess module to execute external system utilities (pdfinfo and pdffonts) to analyze PDF metadata. While these are legitimate calls for document validation, they involve spawning subprocesses using user-supplied file paths.
Evidence: Calls to subprocess.run(['pdfinfo', ...]) and subprocess.run(['pdffonts', ...]) in scripts/validate_format.py.
[INDIRECT_PROMPT_INJECTION]: The skill takes untrusted user input and interpolates it directly into LaTeX source files, creating a potential vector for document-based injection attacks if the generated LaTeX is compiled in an environment with high privileges or shell-escape enabled.
Ingestion points: The --title, --authors, --affiliations, and --email arguments in scripts/customize_template.py.
Boundary markers: None. The script does not use delimiters or provide instructions to the agent to ignore embedded commands in the user-supplied strings.
Capability inventory: SKILL.md provides explicit instructions for the agent or user to compile the resulting files using pdflatex or latexmk.
Sanitization: None. The script uses re.sub to place raw user input into the template without escaping LaTeX control characters.
[METADATA_POISONING]: The documentation in SKILL.md refers to a capability and a specific script (scripts/generate_schematic.py) that are not present in the skill's file list.
Evidence: The 'Visual Enhancement with Scientific Schematics' section in SKILL.md describes and provides command examples for a non-existent script.

venue-templates