# result-poisoning
4 articlestagged with “result-poisoning”
Function Result Poisoning (Agentic Exploitation)
Techniques for manipulating function return values to influence LLM behavior, inject instructions via tool results, and chain poisoned results into multi-step exploitation.
Orchestrator Compromise Techniques
Advanced techniques for compromising central orchestration agents in multi-agent systems, including result poisoning, recursive task bombs, capability confusion, consensus manipulation, fallback path exploitation, and orchestrator prompt injection.
函式結果投毒（代理式攻擊）
操弄函式回傳值以影響 LLM 行為的技術、透過工具結果注入指令，以及將被投毒結果串接為多步攻擊。
Orchestrator Compromise Techniques
進階 techniques for compromising central orchestration agents in multi-agent systems, including result poisoning, recursive task bombs, capability confusion, consensus manipulation, fallback path exploitation, and orchestrator prompt injection.