The Agent Skills Directory

[SAFE]: The skill operates within a well-defined domain of academic writing, using standard system tools like LaTeX, Pandoc, and Quarto for document compilation as part of its core functionality.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard CLI tools, including Git (for diffing and status checks) and various document engines (latexmk, pdflatex, quarto), to support the editing and verification workflow (found in SKILL.md and references/refactor-and-compile.md).
[PROMPT_INJECTION]: The skill defines specific inline markers such as 'TODO', 'XXX', and intent comments as legitimate directives for the agent to identify and complete work, which is a core feature for its drafting and polishing modes (found in SKILL.md and references/draft.md).
[SAFE]: The skill implements robust architectural boundaries through its 'Mode routing' system, which ensures the agent only loads contextually relevant instructions and authority grants based on the specific task (found in SKILL.md and CLAUDE.md).
[SAFE]: The instructions include defensive measures to prevent the leakage of internal project metadata—such as repository paths, branch names, and internal terminology—into the final document, enhancing the privacy and professional quality of the output (found in references/style.md).

writing