research-lit

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

• Third-party content exposure detected (high risk: 0.78). SKILL.md describes runtime ingestion of outsider-authored free text from external sources—specifically WebSearch/arXiv/Semantic Scholar results and Zotero/Obsidian notes/annotations (via MCP tools) that the operating user did not author—into the agent’s LLM context during “Step 1: Search (external)” and “Step 0a/0b” compilation of extracted annotations/summaries.