The Agent Skills Directory

[PROMPT_INJECTION]: The skill extracts text from untrusted .docx files via 'pandoc' for agent processing, which creates a surface for indirect prompt injection. Instructions embedded in a document could influence the AI's behavior.
Ingestion points: SKILL.md (Text extraction section using pandoc).
Boundary markers: Absent. The skill instructions do not provide delimiters or warnings for the agent to treat extracted content as data only.
Capability inventory: The skill uses subprocess.run (in pack.py and redlining.py) and has file system write access.
Sanitization: While defusedxml is used for XML parsing, content extracted via pandoc is processed as raw markdown.
[COMMAND_EXECUTION]: The SKILL.md file suggests using sudo apt-get install commands to install system dependencies. Additionally, internal Python scripts (ooxml/scripts/pack.py, ooxml/scripts/validation/redlining.py) use subprocess.run to execute external tools like 'soffice' and 'git'.
[EXTERNAL_DOWNLOADS]: The documentation encourages downloading and installing third-party tools and libraries from well-known repositories (npm, PyPI, and apt), including 'docx', 'defusedxml', 'pandoc', and 'libreoffice'.

docx