architect
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external codebase subsystems and grounding artifacts during the design phase. These inputs are untrusted and could contain hidden instructions targeting the design synthesis or implementation phases.
- Ingestion points: Grounding artifacts and existing subsystems referenced in SKILL.md (Phase A and Phase B).
- Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the ingested codebase or grounding artifacts.
- Capability inventory: Orchestrates parallel sub-agents (via the 'arena' skill) to generate design packages, rationales, and code skeletons (SKILL.md, Phase B and Phase D).
- Sanitization: There is no evidence of sanitization, escaping, or validation of the content retrieved from external subsystems before it is used to ground the model's design decisions.
Audit Metadata