architect
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill establishes a set of best practices and logical phases for AI-assisted code architecture. It does not attempt to bypass safety guidelines, exfiltrate data, or execute unauthorized code.
- [INDIRECT_PROMPT_INJECTION]: The skill processes existing codebase information during its 'Grounding' phase to understand system integration. * Ingestion points: Subsystems and project files are read in SKILL.md (Phase A) and passed to subsequent phases. * Boundary markers: The skill provides a clear internal structure for processing this information but does not use explicit delimiters or specific 'ignore embedded instructions' warnings for the external code being analyzed. * Capability inventory: The skill generates architectural sketches and implementation code, and invokes other functional skills for specific tasks (how, why, arena). * Sanitization: While no specific sanitization steps are listed, the skill's structure focuses on high-level design rather than direct execution of input data.
Audit Metadata