The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/verify_docx.py uses the subprocess.run function to execute the pandoc utility. This is used to convert Word documents (.docx) into Markdown for automated checking. The command is constructed using a list of arguments, which is a secure practice that prevents shell injection attacks.
[EXTERNAL_DOWNLOADS]: The documentation and scripts/verify_docx.py reference the need for pandoc, an external document conversion tool. The skill itself does not download the binary, but provides a link to the official project website (pandoc.org) for installation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes user-provided report drafts and evaluate them against specific rules.
Ingestion points: The scripts evaluate_report.py and verify_docx.py read content from user-specified file paths and directories.
Boundary markers: The scripts rely on regular expression matching for headings and word counts but do not implement specific delimiters or 'ignore' instructions to isolate user content from the agent's evaluation logic.
Capability inventory: The skill has the ability to read local files, create new markdown files (scripts/generate_shell.py), and call the local pandoc binary.
Sanitization: No input sanitization is performed on the text extracted from documents before it is processed by the evaluation logic.

tim-project-guide