architect
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external subsystems via the
howandwhyskills during its grounding phase. This content is used to inform architectural decisions, creating a potential surface for indirect prompt injection where malicious content in the codebase could influence the agent's logic.\n - Ingestion points: External code and subsystems are processed in Phase A via the
howskill.\n - Boundary markers: The skill uses markdown templates for its rationales but lacks explicit markers to differentiate and isolate ingested code from its own instructions.\n
- Capability inventory: The skill possesses the ability to spawn sub-agents using the
spawn_agenttool and generate implementation code against designs.\n - Sanitization: No explicit sanitization or filtering of input from the subsystems is performed.
Audit Metadata