nextjs-reproduce-issue

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md:8-14 directs the agent to read an outsider-authored “report” (including comments/repository files/web pages) as input and extract expected/actual behavior and steps, which is free text authored by someone other than the operating user and would be placed into the agent’s LLM context during the workflow.