architecture-and-design
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is primarily instructional, providing best practices for clean architecture, SOLID principles, and design patterns. It includes a dedicated security ruleset that advises on preventing XSS via DOMPurify, avoiding hardcoded secrets in client bundles, implementing Content Security Policies (CSP), and validating redirect URLs. These guidelines serve to improve the security posture of the code the agent generates or reviews.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data in the form of code diffs and pull request content (specified in the 'Review' mode in SKILL.md and demonstrated in references/worked-example.md). While this represents a vulnerability surface, the skill includes behavioral constraints such as 'Do not invent findings' and 'One kind of change per commit' which act as guardrails. Furthermore, it advocates for strict input validation ('External input enters as unknown and is narrowed at the boundary') and the use of sanitization libraries, mitigating the impact of potentially malicious content within the analyzed data.
Audit Metadata