agent-safety
AI agent safety standards — fail-closed defaults, least-privilege tool access, human-in-the-loop for destructive actions, prompt injection defense, budget controls, OWASP LLM Top 10 alignment, and governance audit trails for all agent workflows.
Overview
| Property | Value |
|---|---|
| Type | Instruction |
| File | instructions/agent-safety.instructions.md |
| Applies To | **/*.agent.md, **/*.instructions.md |
| WAF Alignment | security, responsible-ai, reliability |
| Lines | 178 |
How It Works
Instructions are automatically applied to files matching the applyTo glob pattern. When a developer opens a matching file in VS Code with GitHub Copilot, this instruction's content is injected into the AI context.
Source
Auto-generated from the FrootAI primitive catalog.