What problem does it solve? Building an Agno agent without input validation leaves it exposed to prompt injection, PII leakage, off-topic requests, and policy violations. This Skill enforces a non-negotiable rule: every agent must run with at least one guardrail configured as a pre-hook, and it provides the exact patterns to do so. ## Core Features & Use Cases - Mandatory baseline stack: Wires PII Detection, Prompt Injection, and OpenAI Moderation guardrails into every agent via pre_hooks, ordered from cheap deterministic checks to external provider calls. - Custom scope guardrail: Provides a ScopeGuardrail extending BaseGuardrail with both check and async_check, blocking out-of-domain messages with InputCheckError and safe refusal messages. - Optional AWS Bedrock Guardrails: Adds a Bedrock ApplyGuardrail integration enabled only via environment variables, with fail-fast validation when enabled without a guardrail ID. - Use Case: When creating or reviewing any Agno agent, apply this Skill to assemble build_guardrails() in the composition root so the agent can never start with an empty guardrail list. ## Quick Start Ask the AI to add the mandatory guardrail stack with a custom scope guardrail to your Agno agent following the agno-guardrails skill.
