← All concepts

prompt injection attack surface

2 articles · 10 co-occurring · 0 contradictions · 0 briefs

Article explicitly calls out 'external content manipulating LLM, polluting HEARTBEAT.md'—this is prompt injection via unvalidated context sources

Related concepts
tool use security 1 tool integration patterns 1 session state integrity 1 permission access control 1 multi agent orchestration 1 model context protocol 1 mcp server integration patterns 1 credential lifecycle management 1 context validation trust boundaries 1 context isolation boundaries 1
Evidence chain (2 articles, showing 2)
@shao__meng: 各位关注和在负责企业级 Agent 研发、部署和安全的朋友们，推荐阅读！ example_of

Article explicitly calls out 'external content manipulating LLM, polluting HEARTBEAT.md'—this is prompt injection via unvalidated context sources

New Prompt Injection Attack Vectors Through MCP Sampling supports

Demonstrates new attack vectors through MCP sampling—expands understanding of where prompt injection threats emerge in multi-tool systems.

query this concept
$ db.articles("prompt-injection-attack-surface")
$ db.cooccurrence("prompt-injection-attack-surface")
$ db.contradictions("prompt-injection-attack-surface")