linkfox-amazon-ads-sp-insights-report

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The entry scripts get_sp_audience_report.py and get_sp_search_impression_share.py execute scripts/check_auth_dependency.py using subprocess.run. This is a deterministic local call used to verify that the required linkfox-amazon-ads-auth skill is available in the environment.
  • [EXTERNAL_DOWNLOADS]: The skill retrieves CSV and Gzip-compressed report data from remote URLs as part of the reporting_v1_workflow.py. These downloads are triggered by Amazon Ads API responses (proxied via the vendor gateway) and are essential for the skill's primary reporting functionality.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from Amazon Ads reporting APIs and provides a summary to the agent, creating a surface for indirect prompt injection.
  • Ingestion points: Data is downloaded from external URLs and read via csv.DictReader in reporting_v1_workflow.py.
  • Boundary markers: The skill does not implement explicit delimiters or instructions to ignore embedded content within the CSV data summarized for the user.
  • Capability inventory: The skill possesses network access to vendor-controlled domains and file-writing capabilities to the local session directory.
  • Sanitization: The skill redacts presigned URLs and performs basic validation of the CSV data structure to ensure integrity during processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 12:42 AM
Security Audit — agent-trust-hub — linkfox-amazon-ads-sp-insights-report