The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by processing external document data.
Ingestion points: External .docx files are read using python-docx (e.g., input.docx) and templates are processed with docxtpl (e.g., template.docx).
Boundary markers: None identified; the skill does not wrap extracted content in markers or provide instructions to ignore embedded commands.
Capability inventory: The skill can read/write files and extract text/tables.
Sanitization: No sanitization or validation of the contents of the .docx files is performed before the data is handled.
[EXTERNAL_DOWNLOADS]: The skill references several standard document processing libraries from public registries.
Python: python-docx, docxtpl.
Node.js: mammoth, docx.

document-docx