Enforce deterministic governance policies on Foundry hosted agent tool actions.
npx skills add https://github.com/aiappsgbb/awesome-gbb --skill foundry-agt
Please help me install this Agent Skill. Skill: foundry-agt Source: https://github.com/aiappsgbb/awesome-gbb/tree/main/skills/foundry-agt Command: npx skills add https://github.com/aiappsgbb/awesome-gbb --skill foundry-agt
This Skill prevents Foundry hosted agents from performing unsafe or noncompliant actions by enforcing deterministic, pre-execution policy checks outside the LLM.
Use the foundry-agt references and policies by wiring the build_governed_agent factory into your Foundry hosted-agent module and loading the references/policies directory for deterministic allow/deny enforcement.
High-intent search queries and answers about installing and using this skill.
To enforce deterministic policy checks on Foundry agent tool actions, you can wrap Microsoft Agent Governance Toolkit middleware around agent-framework agents using pre-loaded YAML rules for allow or deny evaluations before execution.
Preventing unsafe or destructive operations requires applying capability guardrails that explicitly deny specific capabilities, ensuring the agent cannot call destructive tools outside of pre-execution policy evaluations.
Generating tamper-evident audit logs for MCP-triggered tool usage involves producing hash-chained audit records during execution, which can then be exported using CloudEvents format for compliance workflows.
Yes, you can apply human-in-the-loop gating and PII guardrails to Foundry agents by loading starter governance policy files that explicitly define HITL gates and deny rules for sensitive data processing.
Yes, this governance approach evaluates policy rules entirely outside the LLM by utilizing deterministic middleware to intercept and validate tool actions before they are executed, ensuring noncompliant actions are blocked.
vercelRuns untrusted code in ephemeral Firecracker microVMs with network and I/O isolation.
Kgan01Create hash-chained audit logs with genesis entries and prev_hash links.
bmarcurellaAutomate building, provisioning, and deploying AI agents on Microsoft Frontier Agent Platform.
sandgardenhqDesign and register OpenCode plugins with event hooks and custom tools.
