newsletter
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches external content from live pages for research, which exposes the agent to potential indirect prompt injection attacks.
- Ingestion points: Phase 3 uses tools such as Exa, Firecrawl, and WebSearch to read content from external URLs.
- Boundary markers: No specific delimiters or safety instructions are defined to isolate the fetched content.
- Capability inventory: The skill can write to the local filesystem and make network requests to ESP APIs.
- Sanitization: Uses Firecrawl for cleaning text but lacks explicit prompt-level sanitization for malicious instructions.
- [COMMAND_EXECUTION]: The skill executes a Python script from a local directory for brand context setup.
- Evidence: It calls
python3to runbrand_context.pyduring initialization. - [EXTERNAL_DOWNLOADS]: The skill performs network requests to external APIs and web pages.
- Evidence: It connects to Beehiiv, Kit, and Ghost APIs to retrieve newsletter performance data.
Audit Metadata