write-outreach
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of prospect replies and outbound campaign statistics.\n
- Ingestion points: Prospect replies are pasted into the context or read from external files (list.csv, results exports) to generate draft responses.\n
- Boundary markers: The skill utilizes specialized sub-skills (write-outreach-reply-reader) to classify and process replies, although explicit text delimiters for data interpolation are not defined in the instructions.\n
- Capability inventory: The skill has permissions to write files (send.csv, pack.md), read brand knowledge, and interact with the filesystem via MCP tools.\n
- Sanitization: All drafted content is routed through a dedicated QA skill (write-outreach-outreach-qa) and checked against compliance rules and character limits before delivery to the user.
Audit Metadata