design-exploration
Fail
Audited by Socket on Mar 18, 2026
1 alert found:
MalwareMalwareSKILL.md
HIGHMalwareHIGH
SKILL.md
SUSPICIOUS: the overall workflow fits a design-exploration purpose, but it relies on multiple external agent systems and a mandatory Kimi MCP delegation path that is not verifiably official from the provided evidence. The main risk is external trust and data flow to unverified MCP tooling, plus indirect prompt-injection exposure from web/browser research combined with file-writing and agent actions.
Confidence: 82%Severity: 71%
Audit Metadata