product-diagnosis
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted external data from multiple sources, creating a surface for indirect prompt injection. Malicious instructions embedded in customer feedback or Slack messages could be processed and followed by the agent.\n
- Ingestion points: Data is retrieved from customer feedback (
get_feedback_mentions,get_feedback_comments), AI agent conversation transcripts (get_ai_conversation,search_ai_conversations), and Slack searches (slack_search_public) as detailed in Phase 2d, 2e, and 4 of SKILL.md.\n - Boundary markers: The instructions lack requirements for delimiters or specific warnings to ignore instructions embedded within the ingested data.\n
- Capability inventory: The skill uses tools for data searching, retrieval, and analysis. While no direct shell execution or file-writing tools are explicitly defined in the instructions, the agent processes this data to generate recommendations and potentially interact with the user.\n
- Sanitization: There is no evidence of sanitization or validation of the retrieved text content before it is passed to the language model for synthesis.
Audit Metadata