The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands involving pandoc, bash, and python3, including the local scripts scripts/convert-to-pdf.sh and scripts/fix-ascii-art.py for document processing.
[COMMAND_EXECUTION]: Workflow instructions guide the agent to construct shell commands using user-provided file paths. This creates a surface for command injection if paths are not properly quoted, though the skill provides explicit warnings to the agent to quote all paths and verify file existence.
[DYNAMIC_EXECUTION]: The scripts/convert-to-pdf.sh script dynamically generates a temporary LaTeX header file at runtime using mktemp and populates it with formatting directives before use in the conversion process.
[DYNAMIC_EXECUTION]: Documentation in README-zh.md suggests using shell process substitution (<()) to dynamically provide LaTeX headers to pandoc, which involves executing a subshell command string.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documents (Markdown, Word, HTML) which may contain malicious pandoc extensions, Lua filters, or LaTeX engine commands. If processed without strict restrictions, these files could trigger unauthorized file system access or code execution through the pandoc toolchain.
[EXTERNAL_DOWNLOADS]: Documentation references external fonts and citation styles from well-known repositories on GitHub (such as adobe-fonts and the citation-style-language project) for environment configuration.

pandoc-converter