The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided Word and Markdown documents, which establishes a surface for indirect prompt injection where malicious instructions embedded in a document could attempt to influence the agent's behavior.
Ingestion points: scripts/docx_reader.py (reads .docx files) and scripts/md_to_docx.py (reads .md files).
Boundary markers: The Markdown representation of document content used for AI analysis lacks explicit security delimiters or warnings to separate document content from agent instructions.
Capability inventory: The scripts/docx_editor.py utility can modify, insert, or delete paragraphs, tables, and document metadata.
Sanitization: No text sanitization or instruction filtering is performed on data ingested from documents.
[SAFE]: All document manipulations are performed locally on the file system. No network operations, hardcoded credentials, or obfuscated code patterns were found. The skill also includes a safety feature in scripts/docx_editor.py that automatically creates backups (.bak files) before modifying any document.

word-master