agents-system
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill acts as a metaprogramming guide for repository analysis and redesign. It respects user-defined boundaries, includes a dry-run option, and explicitly restricts runtime changes to those authorized by the user. Its operations are focused on documentation and implementation planning within the local file system.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface through the ingestion of external repository content. Ingestion points: the agent reads repository entry points, manifests, and representative code (SKILL.md, Section 1). Boundary markers: no explicit markers are defined to isolate ingested data from agent instructions. Capability inventory: the agent is authorized to update design documents and implementation plans (SKILL.md, Section 4). Sanitization: no specific sanitization methods for ingested content are described. This finding represents an inherent surface for analysis tools and is categorized as low risk given the output is restricted to documentation.
Audit Metadata