agent-first-pages
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were identified in the skill instructions. The skill is primarily instructional and focused on critique logic.
- [DATA_EXFILTRATION]: While the skill mentions the ability to fetch project files via MCP (Model Context Protocol), it is explicitly defined as a read-only interaction and does not include logic for transmitting data to external third-party servers.
- [PROMPT_INJECTION]: The skill includes explicit instructions for the agent to 'hold its ground' and resist user pressure or authority-style rebuttals, which acts as a safeguard against sycophancy and certain types of behavioral manipulation.
Audit Metadata