amazon-sp-refund-return-monitor
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external Amazon return reports that contain untrusted data in the form of customer comments.
- Ingestion points: Data is ingested through user-provided files (TSV/CSV/JSON) or live SP-API report retrieval as described in
SKILL.md(Step 3) andreferences/sp-api-data-sources.md. - Boundary markers:
SKILL.mdprovides explicit instructions to "Minimize, redact, and avoid reproducing raw rows" and "Do not paste raw files or wide row sets into chat". - Capability inventory: The skill is designed for data analysis and does not include instructions for arbitrary command execution or external network writes; it relies on the host's native tools for data processing and SP-API access.
- Sanitization: The instructions in
SKILL.mdandreferences/column-mapping-hints.mdemphasize redacting sensitive personal information (PII) like names and addresses, and using redacted paraphrases for customer comments.
Audit Metadata