system-architect
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates as a purely instructional guide for generating documentation (ARCHITECTURE.md, DATABASE.md, SECURITY.md) based on an input file (PRD.md). It does not contain executable code, scripts, or commands.
- [SAFE]: Includes a 'Greenfield Overwrite Guard' that explicitly prevents the agent from overwriting existing specification files without user confirmation, demonstrating defensive design.
- [SAFE]: The instruction to verify library versions using 'available documentation/search tools' is a standard use of an agent's search capabilities to ensure technical accuracy and does not involve downloading untrusted executable content.
- [SAFE]: The skill instructs the agent to define security contracts including secrets registries, tenant isolation, and prompt-injection sanitization rules for the target system, which aligns with security best practices.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from
specs/PRD.md. While this presents a theoretical attack surface where a malicious PRD could influence the generated architecture, the skill's role is restricted to document generation and lacks dangerous execution capabilities (like shell access or network exfiltration), making the risk negligible.
Audit Metadata