The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external source (Google Calendar events) which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The list_events, get_event, and search_events functions in scripts/calendar_operations.py retrieve event summary, description, and location data from the Google Calendar API.
Boundary markers: While SKILL.md mandates user confirmation for destructive actions, there are no technical delimiters or instructions to ignore embedded commands when processing event data.
Capability inventory: The skill possesses the ability to create, update, and delete calendar events and manage attendees via the create_event, update_event, delete_event, and attendee management functions in scripts/calendar_operations.py.
Sanitization: No sanitization, escaping, or filtering of the retrieved event content is performed before it is presented to or processed by the agent.
[DYNAMIC_EXECUTION]: The Python script scripts/calendar_operations.py dynamically modifies the system path using sys.path.insert to load shared authentication modules from a computed location relative to the project root. This is used for internal module resolution to share authentication logic between related Google skills.
[COMMAND_EXECUTION]: The skill executes a local Python script to interact with the Google Calendar API. This is the intended primary functionality and includes manual confirmation prompts for all modifications, such as creating, updating, or deleting events.

google-calendar