The Agent Skills Directory

[COMMAND_EXECUTION]: The module scripts/core/compiler.py utilizes subprocess.run to execute LaTeX compilation sequences. While the commands executed are configurable via the pass_sequence field, the skill's logic is designed to invoke standard academic tools like xelatex, bibtex, and biber to process project files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from various file formats (LaTeX, Word, PDF, Markdown). As the AI agent is instructed to summarize, optimize, or adapt this content, it is exposed to potential indirect prompt injection embedded within the source documents. The skill attempts to mitigate disruption to functional code via the ReferenceGuardian component, which replaces LaTeX commands with unique placeholders during AI processing.
[SAFE]: The skill implements a robust SecurityManager that validates all file write operations against a whitelist defined in config.yaml. This prevents the skill from modifying sensitive system or template files (e.g., .cls, .sty, or main.tex), ensuring that the tool only impacts the intended content layers.

transfer-old-latex-to-new