gws
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the
@anthropic-ai/gwscommand-line utility from the npm registry. This tool is an official package provided by Anthropic for interacting with Google Workspace services. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The agent can ingest untrusted data when reading email bodies via
gws gmail users messages getor viewing calendar event descriptions viagws calendar events listandgws calendar +agenda. - Boundary markers: The instructions do not explicitly mandate the use of delimiters (like XML tags or triple quotes) when the agent processes retrieved content, though it encourages specific search queries.
- Capability inventory: The skill possesses write/delete capabilities, including sending emails (
gws gmail +send), replying to threads (gws gmail +reply), trashing messages (gws gmail users messages trash), and deleting calendar events (gws calendar events delete). - Sanitization: The skill provides a significant mitigation by explicitly instructing the agent to 'Always ask for confirmation before sending emails or creating/modifying/deleting calendar events,' which places a human-in-the-loop to prevent automated execution of malicious instructions found in data.
Audit Metadata