The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides comprehensive instructions for executing shell commands via the gog CLI to interact with Google services including Gmail, Calendar, Drive, and Sheets.
[DATA_EXFILTRATION]: The skill has significant data access and transmission capabilities.
Data Access: It can read sensitive content from Gmail messages (gog gmail get), Google Drive files (gog drive download), and Google Docs (gog docs cat).
Exfiltration Channels: It can send data externally via email (gog gmail send) or by creating publicly accessible sharing links for private files (gog drive share --role=reader --type=anyone).
[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection attacks.
Ingestion Points: The agent is instructed to read content from untrusted external sources, specifically the bodies of unread emails (gog gmail search) and the contents of documents in Google Drive (gog docs cat).
Capability Inventory: The agent has high-privilege capabilities including sending emails, deleting files, and modifying calendar events across the user's entire Google Workspace.
Boundary Markers: There are no explicit instructions or delimiters provided to the agent to help it distinguish between its primary instructions and potentially malicious commands embedded within the processed data.
Sanitization: The skill does not implement any validation or filtering of the content ingested from external messages or files before processing.

google