AGSAFEstill growing

Safety and guardrails for agents

Name the risks of an agent that reads text and calls tools, and put the right control in front of each one.

For Developers moving from a chatbot to an agent that can act, and who want to know what can go wrong.

PrerequisitesAGENTS Agents from first principles

7 notes, in order. Each one assumes the one before it. Still growing: more notes may be added.

Notes in this trail

Start the trail
  1. Probabilistic risks, not yet read
  2. LLM guardrails, not yet read
  3. Hallucination prevention, not yet read
  4. Human in the loop, not yet read
  5. Ejection logic, not yet read
  6. Secure tooling, not yet read
  7. Agent ethics, not yet read