team-design
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell and Node.js scripts including
discover-topic.sh,design-review-status.sh, andrecord-design-review.mjs. These scripts are called with the$ARGUMENTSdirectory path via shell interpolation. While the variables are double-quoted, this practice maintains an attack surface for path-based command manipulation. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by ingesting external data from
5-research.mdand6-design.md(Ingestion points:SKILL.md). It relays reviewer reports and findings verbatim to agents without using boundary markers or sanitization (Boundary markers: absent; Sanitization: absent). Malicious content within the ingested artifacts could potentially manipulate thedesign-authoragent, which utilizes capabilities such as theWritetool and various script executions (Capability inventory:node, shell scripts,Exploresubagent).
Audit Metadata