the-subject-line-lab
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting and processing untrusted user data (the email body and target persona).
- Ingestion points: The agent asks the user for a full email body, target persona, and main hook via direct input.
- Boundary markers: The skill does not implement specific delimiters or 'ignore embedded instructions' warnings for the pasted email body text.
- Capability inventory: The skill is restricted to text generation; it lacks the ability to execute shell commands, perform network requests, or write to the file system.
- Sanitization: No validation or filtering is performed on the user-supplied email body content.
- [SAFE]: The skill reads from local configuration files such as
.agents/product-context.mdand uses standard reference guidelines inreferences/outbound-copy-standards.md, which is a standard pattern for providing task-specific context to an agent. - [SAFE]: No hardcoded credentials, malicious obfuscation, or remote code execution patterns were detected in the skill scripts or instructions.
Audit Metadata