connected-service-automation
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is primarily a set of defensive instructions aimed at ensuring operational security when interacting with connected services. It explicitly forbids the exposure of credentials, tokens, or sensitive configurations (in
references/accounts-and-capabilities.md) and mandates the use of bounded retries and idempotency to prevent accidental data duplication (inreferences/safe-mutations.md). - [INDIRECT_PROMPT_INJECTION]: The skill interacts with external, user-controlled data (e.g., messages, files, and notes), which creates a potential surface for indirect prompt injection attacks.
- Ingestion points: The skill reads external data such as conversation history and file contents as described in
references/messaging-and-sharing.mdandreferences/records-and-media.md. - Boundary markers: The skill instructions prioritize human-in-the-loop verification (e.g., "Show the final message before sending") and state checks over automated text delimiters.
- Capability inventory: The skill is intended to operate tools with the ability to send messages, modify files, and change resource permissions.
- Sanitization: Explicit input sanitization or filtering is not detailed; instead, the skill focuses on verifying the stability of identifiers and intended state changes.
Audit Metadata