09-insight-khach-hang
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted data from external sources provided by the user.
- Ingestion points: In the 'Thu thap thong tin' section (Question 3) and 'Quy trinh thu thap insight tu review/comment that' (Step 1), the agent is instructed to ask for and process CRM data, customer inbox/messages, and social media reviews/comments.
- Boundary markers: Absent. There are no instructions for the agent to use delimiters or ignore embedded commands when processing the provided customer feedback.
- Capability inventory: The skill is strictly instructional and generates markdown output. No subprocess calls, network requests, or file-writing scripts are present.
- Sanitization: Absent. The skill does not provide mechanisms for filtering or sanitizing the ingested customer language bank or message data.
Audit Metadata