The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/Code.js contains a hardcoded default fallback for the SECRET_TOKEN variable ('aviz-gmail-2026'). If a user fails to correctly execute the setupToken() function or if the environment variable is lost, the web endpoint remains protected by this publicly visible string.
[DATA_EXFILTRATION]: The skill is designed to transmit sensitive Gmail data (email bodies, sender info, and subjects) to a remote web app endpoint. While this is the skill's primary function, the exposure of private communication to an external endpoint increases the risk of data leakage.
[EXTERNAL_DOWNLOADS]: The SETUP.md file instructs the user to install @google/clasp globally via NPM. This is a tool from a trusted organization (Google) used for managing script deployments.
[PROMPT_INJECTION]: The skill reads external, untrusted email content which may contain malicious instructions that could influence the agent's behavior during processing (Indirect Prompt Injection).
Ingestion points: Raw email bodies and snippets are retrieved in scripts/Code.js using GmailApp.search and msg.getPlainBody().
Boundary markers: None; the content is returned as raw text without delimiters or warnings to the agent.
Capability inventory: The skill provides capabilities to send emails, create drafts, and modify message states (mark as read).
Sanitization: No validation or sanitization is performed on the ingested email content.

gmail