概念导读
Agent safety covers methods for preventing tool-using and autonomous agents from executing harmful, unsafe, or policy-violating behavior across planning, memory, tool use, and multi-step execution.
- Related
Agent Safety
Agent safety covers methods for preventing tool-using and autonomous agents from executing harmful, unsafe, or policy-violating behavior across planning, memory, tool use, and multi-step execution.
In this wiki it serves as an umbrella page linking training-time safety, runtime defense, memory safety, and prompt-injection risks.
Related
- agent-safety-via-rl — RL-based safety training for agents.
- thought-aligner — Runtime thought correction for unsafe reasoning.
- openclaw-safety-analysis — Real-world safety analysis for computer-use agents.
- prompt-injection — Attack class against tool-using agents.