email-creative-builder
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, exfiltration attempts, or unauthorized command executions were found in the skill files.
- [SAFE]: The skill incorporates an attack surface for indirect prompt injection by processing external data such as campaign CSVs, scraped landing page copy, and pasted emails. However, it explicitly includes security instructions to treat these as untrusted inputs and disregard any embedded commands, which is a security best practice.
- Ingestion points: CSV exports, landing-page scraping, pasted competitor emails, and CRM signals.
- Boundary markers: Explicit instructional warnings are present in the 'Instructions' section.
- Capability inventory: Limited to internal state management via reading and writing to 'memory/' directory paths; no external network or arbitrary code execution capabilities detected.
- Sanitization: Relies on specific instructions to ignore embedded content within untrusted data and refers to a global SECURITY.md policy.
- [SAFE]: External references are limited to the author's own GitHub repository and standard internal skill references within the agent's environment.
Audit Metadata