Self-hosted alternatives to agent-governance

A

agent-governance

Declarative policies, intent classification

agent-governance is an agent skill from the github/awesome-copilot collection that describes patterns for adding governance, safety and trust controls to AI agent systems. The core flow runs from a user request through intent classification and a policy check, then tool execution and an audit log, with threat detection and allow or deny decisions along the way. The skill's central idea is that policies are declarative: they are stored as configuration rather than code, so they can be reviewed and changed without rewriting the agent. Composable policies can define which tools are allowed or blocked, which content filters apply, what rate limits hold, and which actions require approval from a person. Audit trails record what the agent did and why, so the actions can be explained afterwards. The skill applies when building agents that call external tools or act on behalf of users, and when deciding how much an agent should be allowed to do without oversight. Governance patterns reduce risk but do not remove it, so the policies should be tested against realistic requests and reviewed as the agent's tools change. Teams should also keep a human in the loop for actions that are hard to reverse. The skill is installed with the skills CLI from the github/awesome-copilot repository. Its skills.sh page showed about 9.6K installs and passing results from three security audits at the time of the scan.