The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface by ingesting external content from Google Docs.
Ingestion points: Untrusted content enters the agent context via the documents read and documents get commands in scripts/google-docs.py (functions read_document_content, export_document_as_markdown).
Boundary markers: No specific delimiters or instructions to ignore embedded content are applied to the retrieved document text.
Capability inventory: The skill has broad document modification capabilities (append_text, insert_text, delete_content, apply_formatting) and can write to the local file system via PDF export.
Sanitization: No sanitization or filtering is performed on the content retrieved from the Google Docs API.
[COMMAND_EXECUTION]: The skill uses a Python CLI script (scripts/google-docs.py) to interface with Google APIs, which is invoked by the agent via Bash as defined in the allowed-tools metadata.
[EXTERNAL_DOWNLOADS]: The documentation guides the user to install standard dependencies from the official Python Package Index (PyPI), including google-auth, google-api-python-client, and keyring for secure secret management.
[DATA_EXFILTRATION]: The skill performs network operations to googleapis.com and accounts.google.com. These operations are limited to the intended purpose of document management and standard OAuth 2.0 authentication flows.

google-docs