“llm-security”
Delegated Authorities: Scoped Permissions for Sub-Agent Execution
Give sub-agents only the authority their task requires: why tool lists are not permissions, how agents-as-tools and handoffs differ in privilege, building a delegation layer with capability objects, and enforcing scope at the runner rather than trusting the prompt.
Consent Flows: Human Approval at the Point of Action
Design consent flows that pause an agent before it acts: deciding which tools need approval, the pause-and-resume state machine, sticky decisions, rejection messages, server-side approval security, and how to build the same pattern on the Gemini function calling loop.
Toxicity and Safety Filters for LLM Outputs: A Hands-On Guide
Implement production-ready toxicity and safety filters for LLM outputs. Classify, filter, and sanitize harmful content before it reaches your users.
Building an Agent Risk Register from the OWASP LLM Top 10
How to build a practical agent risk register using the OWASP LLM Top 10 and Agentic Top 10 — with risk scoring, mitigation strategies, and code examples.