Sort controls by layer
Classify prompt-security mitigations by the layer where they enforce risk reduction.
Sort each mitigation into the control layer where it primarily works. Prompt guidance Runtime validation Permission design Human/process control System prompt says external documents are evidence, not instructions Output validator rejects answers without required claim-source fields Tool uses a user-scoped read-only OAuth token Finance owner must approve payment emails before send Retriever filters sources by tenant and sensitivity before ranking Agent has no delete function exposed in its tool list Release checklist requires three adversarial prompt-injection tests Developer instruction requires the model to say not found when sources are missing Prompt Runtime Permission Process
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in