team-communications
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions create a significant surface for indirect prompt injection by ingesting data from untrusted external sources.
- Ingestion points: According to
references/3p-updates.md,references/company-newsletter.md, andreferences/faq-answers.md, the agent is instructed to pull information from Slack channels, Gmail threads, Google Drive documents, Calendar events, and external press articles. - Boundary markers: The skill lacks explicit instructions to use delimiters or protective markers when interpolating external content into the agent's context. There are no directives for the agent to ignore or neutralize embedded instructions within the retrieved data.
- Capability inventory: The skill uses MCP tools to read sensitive communication and document stores (Slack, Gmail, Google Drive, Calendar).
- Sanitization: The skill does not implement or require any validation, filtering, or escaping of the ingested content before it is used to generate drafts.
Audit Metadata