datalineage-summary
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- External Resource Access: The skill utilizes the
read_urltool to dynamically fetch location lists from official Google Cloud documentation. This is used to ensure the lineage search covers all relevant regions and relies on established, well-known service domains. - Indirect Prompt Injection Consideration: The skill processes external data retrieved from the Data Lineage API to generate summaries. While this introduces a surface for indirect prompt injection, the skill is designed to transform this data into a structured Markdown report, and the risks are mitigated by the specific context of summarizing data flows.
- Ingestion points: Data enters the agent via the
search_lineageMCP tool and documentation fetched viaread_url. - Boundary markers: The prompt instructions use specific headers (e.g.,
**Upstream Lineage:**) to structure the output, which helps maintain context. - Capability inventory: The skill uses tools for lineage searching and documentation reading; it does not perform file-system writes or arbitrary command execution based on the processed data.
- Sanitization: The instructions guide the agent to prioritize specific asset names and aggregate counts, focusing on a structured summary format.
- MCP Server Configuration: The skill references a Model Context Protocol (MCP) server for lineage traversal. The provided configuration uses official API endpoints and standard authentication methods, which is a common practice for cloud-integrated AI agent skills.
Audit Metadata