The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute build commands, such as latexmk or repository-specific scripts, to compile LaTeX documents. This allows for the execution of local commands as part of the manuscript engineering workflow.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including LaTeX source files, templates, bibliographies, and compilation logs, while maintaining command execution capabilities. This combination presents an indirect prompt injection surface.
Ingestion points: Project tree, templates, bibliography files, figures, and compilation logs in the user's repository.
Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions embedded in the processed files.
Capability inventory: Execution of shell commands for document building and inspection of the project environment.
Sanitization: There is no mention of sanitizing or validating the input files or the build commands extracted from the repository.

latex-writer