insideout
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill functions as a management interface for cloud infrastructure tasks. No malicious patterns, obfuscation, or unauthorized access attempts were identified.
- [NO_CODE]: The skill consists entirely of configuration and instructional markdown without any local scripts or executable code, reducing the local attack surface.
- [DATA_EXFILTRATION]: The skill facilitates the transfer of project summaries to a remote service hosted by the vendor (luthersystems.com). This is an intended functionality for cloud architecture design. The skill includes explicit and robust instructions to prevent the inclusion of sensitive data such as credentials, secrets, PII, or source code in these transmissions.
- [PROMPT_INJECTION]: The instructions direct the agent to act as a transparent relay for the remote service, including a requirement to show responses verbatim. While this overrides standard agent behavior, it is a documented design choice for maintaining the integrity of the expert system's advice and does not involve bypassing safety guidelines.
Audit Metadata