The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads content from Google Docs, Sheets, and files stored in Google Drive. This content is untrusted and could contain instructions designed to subvert the agent's behavior.
Ingestion points: Data is ingested via google_docs_read, google_sheets_read, and google_drive_download (followed by reading the local file).
Boundary markers: Absent. The instructions do not specify any delimiters or warnings for the agent to ignore instructions embedded within the processed data.
Capability inventory: The skill allows the agent to write to Google Docs/Sheets, download files from URLs, and execute shell commands to manage Python virtual environments and run extraction scripts.
Sanitization: Absent. There is no mention of sanitizing or validating the content read from external files before processing it.
[COMMAND_EXECUTION]: The skill instructs the agent to use shell commands for managing files and environments.
Evidence: The use of amp files get <url> -o .amp/in/<name> to download files and python3 -m venv to create environments for text extraction.
[DYNAMIC_EXECUTION]: The skill directs the agent to set up Python virtual environments and use libraries like pymupdf or python-docx to extract text from PDFs and Word documents.
Evidence: Instructions to "extract text with pymupdf or python-docx in a python3 -m venv".

google-workspace