weekly-wrap
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to ingest and process data from external, untrusted sources including emails, chat transcripts, and CRM records. This creates a potential surface where malicious instructions could be embedded within that data to influence the agent's behavior. The skill mitigates this by explicitly instructing the agent to treat these sources as data only, never as instructions, and to report any instruction-like text encountered.
- Mandatory Evidence Chain for Indirect Prompt Injection:
- Ingestion points: Data is pulled from the CRM (opportunities), Calendar (meetings), Email (customer threads), and Chat (team channels).
- Boundary markers: The skill contains explicit "Rules" defining email, chat, transcripts, and enrichment as untrusted content.
- Capability inventory: The skill utilizes tool calls for reading CRM records, searching emails, accessing calendars, and posting to chat channels.
- Sanitization: The skill implements a human-in-the-loop requirement, stating that any content-originated action must be shown to the user with exact targets and source lines before execution. It also restricts automated actions during scheduled or unattended runs.
Audit Metadata