The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/compile.py uses the subprocess module to execute LaTeX engines such as xelatex, lualatex, and pdflatex, as well as latexmk and bibliography processors. This is a core requirement for document generation. The script correctly warns the user about the security implications of the optional --shell-escape flag.\n- [EXTERNAL_DOWNLOADS]: The scripts/online_bib_verify.py script performs HTTP requests to the CrossRef and Semantic Scholar APIs using urllib.request. These requests are used to verify bibliography metadata against trusted academic databases and are consistent with the skill's stated purpose.\n- [COMMAND_EXECUTION]: The scripts/check_format.py script executes the chktex utility via subprocess.run to provide syntax checking for LaTeX files. This is a standard and expected behavior for a LaTeX assistant.\n- [PROMPT_INJECTION]: The skill analyzes user-provided LaTeX and Typst files, which introduces a surface for indirect prompt injection. \n
Ingestion points: Document source code is read by scripts like deai_check.py and scripts/parsers.py. \n
Boundary markers: The SKILL.md provides strict instructions to the agent to output suggestions in a specific LaTeX diff-comment format, which serves to distinguish agent output from user data. \n
Capability inventory: The skill can execute local compilers and access academic APIs. \n
Sanitization: No active sanitization is performed on the input text before processing, but the restricted output format effectively limits the impact of potential instructions hidden in the processed documents.

latex-thesis-zh