The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted LaTeX source files, which creates a potential surface for indirect prompt injection attacks.
Ingestion points: The skill reads user-provided .tex files (e.g., paper/main.tex) through the check and fix actions.
Boundary markers: There are no explicit instructions or delimiters mentioned in the skill definition to prevent the agent from following natural language instructions hidden in LaTeX comments or math environments.
Capability inventory: The skill has the capability to write to the file system (e.g., main_fixed.tex) and execute local Python scripts as defined in the SKILL.md instruction set.
Sanitization: The clean_latex.py script specifically removes zero-width characters (\u200b) and sanitizes special characters for LaTeX compilation safety, but it does not sanitize or filter the content for malicious LLM instructions.

latex-formatting