The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/compile.py utilizes subprocess.run to execute standard LaTeX compilers such as xelatex, lualatex, and latexmk. The skill includes a significant safety feature that prevents the use of the potentially dangerous --shell-escape flag unless the user explicitly provides a --trusted-source confirmation.
[COMMAND_EXECUTION]: The scripts/check_format.py script attempts to run the chktex utility to perform syntax checking if it is installed on the local system.
[EXTERNAL_DOWNLOADS]: The scripts/online_bib_verify.py component performs network GET requests to well-known scholarly databases, specifically api.crossref.org and api.semanticscholar.org, to verify bibliography entries. This functionality is restricted to scholarly metadata retrieval and is intended to be used only with user authorization.
[INDIRECT_PROMPT_INJECTION]: The skill documentation includes a robust security boundary that instructs the agent to treat all user-provided .tex and .bib files as untrusted data. It explicitly commands the agent to ignore any instructions embedded in those files that attempt to reveal prompts, read sensitive files, or override the skill's defined workflow.

latex-thesis-zh