technical-interviewing

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The companion-check section explicitly instructs the agent to run shell commands that list local skill directories and offer installations (ls and npx), which is a clear, out-of-scope instruction that requests access to the host environment and potential exfiltration/side effects unrelated to technical interviewing.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The SKILL.md "Companion check" (On first activation) instructs the agent to run local listing commands and to offer/install missing companion skills via "npx skills add ...", which would fetch untrusted third-party npm packages that the agent could ingest and that might change its behavior.