The Agent Skills Directory

[COMMAND_EXECUTION]: The script executes the latexmk command using subprocess.run to compile the LaTeX project. The execution is performed with -interaction=nonstopmode and -halt-on-error, and it does not enable shell-escape, limiting the risk of malicious LaTeX commands affecting the host system.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data which is subsequently reflected in a report read by the agent.
Ingestion points: Reads latex/main.tex and the resulting main.log (scripts/run.py).
Boundary markers: Absent; the compilation report does not use specific delimiters or warnings to isolate log content.
Capability inventory: Subprocess execution of latexmk, file system write access via shutil.copy2 and atomic_write_text (scripts/run.py).
Sanitization: Absent; the tail of stdout, stderr, and log files are included directly in the markdown report without filtering or escaping.

latex-compile-qa