subject-line-lab
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes explicit security instructions directing the agent to treat all user-supplied content (CSV exports, pasted text, competitor data) as untrusted input and to ignore any instructions embedded within that data. This follows industry best practices for mitigating indirect prompt injection.\n- [DATA_EXFILTRATION]: No network request capabilities, hardcoded credentials, or access to sensitive system files were found. All processing is described as local or heuristic.\n- [COMMAND_EXECUTION]: There are no shell commands, system calls, or use of dynamic context injection (
!command) within the skill files.\n- [REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts or packages.
Audit Metadata