The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the gws command-line utility to interact with the Gmail API, performing operations such as retrieving profiles and managing mailboxes.
[DATA_EXFILTRATION]: The skill is designed to read sensitive email data, including message bodies, headers, and user profiles. This data access is essential for its primary purpose of email management and does not show signs of unauthorized exfiltration.
[PROMPT_INJECTION]: There is an inherent surface for indirect prompt injection as the skill processes untrusted data from incoming emails via commands like +read and +triage.
Ingestion points: SKILL.md (via +read and +triage helper commands that fetch external email content).
Boundary markers: None specified within the skill instructions to delimit external content.
Capability inventory: Sending, forwarding, and replying to emails, as well as managing account settings.
Sanitization: No explicit sanitization or filtering of incoming email text is performed before it is presented to the agent.

gws-gmail