The Agent Skills Directory

[COMMAND_EXECUTION]: The INSTRUCTIONS.md file advises using the -shell-escape flag when compiling LaTeX documents (e.g., uplatex -shell-escape document.tex). This flag allows the compilation process to execute arbitrary shell commands on the host system, creating a significant security risk if the LaTeX source is untrusted or contains malicious code.
[COMMAND_EXECUTION]: In references/DOCUMENT-CLASSES.md, the skill includes an example compilation script and instructs the user to grant it full permissions using chmod 777 compile. Setting world-writable and world-executable permissions on scripts is an insecure practice that facilitates unauthorized modification and execution.
[REMOTE_CODE_EXECUTION]: The -shell-escape flag enables a Remote Code Execution (RCE) vector. Since the skill is designed to process and compile .tex files, an attacker providing a malicious document could gain arbitrary command execution on the agent's environment.
[COMMAND_EXECUTION]: The skill exhibits an indirect command execution surface related to processing untrusted data.
Ingestion points: Processes .tex files as indicated in the skill's description in SKILL.md.
Boundary markers: Absent; no delimiters or warnings are used to differentiate untrusted document content from agent instructions.
Capability inventory: Compilation using uplatex or latexmk with the -shell-escape flag, which allows arbitrary shell commands (documented in INSTRUCTIONS.md).
Sanitization: Absent; the skill does not suggest any escaping or validation of LaTeX source code before compilation.

writing-latex