qmd

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes an MCP/HTTP runtime workflow where the agent sends user-authored search queries (e.g., via the query tool or POST /query body) into QMD, which can then retrieve matching markdown content from the configured local collections; while that retrieved markdown is “inside” the tenant, the workflow ingests free text from outsiders before selecting any specific item.