ad-performance-analysis

Pass

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the bash tool to execute various warehouse and system CLI tools including bq, gcloud, snow, databricks, python3, and open. These are necessary for querying data warehouses, managing local configuration, and opening generated dashboards.
  • [EXTERNAL_DOWNLOADS]: The skill communicates with Fivetran's official API (api.fivetran.com) to retrieve connector and transformation metadata. It also fetches the Chart.js library from cdn.jsdelivr.net to render visualizations in generated dashboards. Both sources are well-known or vendor-owned and are considered safe.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted advertising data (campaign names, keyword text) from external ad platforms via the user's data warehouse. The instructions include guidance for the agent to distinguish between user-defined terms and inferred rules from the data, and the dashboard generator uses HTML escaping for data values.
  • [DYNAMIC_EXECUTION]: The generate-dashboard.py script assembles HTML files that include JavaScript provided in a JSON payload generated by the agent. While this is a form of dynamic execution, it is the primary method for providing interactive visualizations and is limited to local dashboard generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 19, 2026, 06:31 PM