AI Governance Infrastructure — local evaluation. The governance layer for AI agents: a single compiled Rust binary that authenticates, authorises, rate-limits, logs, and costs every AI interaction. Self-hosted, air-gap capable, provider-agnostic.
AGENTS.md instructions for systempromptio/systemprompt-template, covering agents.md — systemprompt-template, what this is, try it, where things live and governance pipeline.
Claude Code instructions for systempromptio/systemprompt-template, covering claude.md, enterprise demo, branching & release flow, building against local core (next tracks next) and quick start.
Read the workspace at a glance: spend, request volume, agent performance, tool reliability, sessions, and conversations - all rolled up over a time window. This is the fleet view; for drilling into one request use inspectairequests, and for one conversation use inspectconversation.
Drive every stage of the governance pipeline end to end, then prove that each decision was audited. This is the guided tour of the enforcement spine: the same four stages run on every tool call in the workspace.
Work with the agents in the workspace: list them, check their configuration and the MCP tools they can reach, send one a message over the A2A protocol, and read the artifacts and traces it produces.
Follow one inference request through the AI gateway: from the client's /v1/messages call, through model routing and the governance pipeline, to the provider and back. Use it to debug a single request - routing, denials, latency, cost, or the tool calls it spawned.
Show the structured data behind a conversation that is happening right now. When a client like Claude Desktop talks to the gateway, every message, AI request, tool call, governance decision, and cost lands in the database keyed to one session. This skill reconstructs that whole picture from a single starting id.
Discover and exercise the MCP plugin surface - especially the admin-only systemprompt MCP server that runs CLI commands as a tool - and keep the skills catalogue in sync.
Change what a user is allowed to do, then watch the governance pipeline honour the change on the very next call - no restart, no new token. This is the live proof that authority is data, not a baked-in constant.
A deliberately dangerous capability that the platform refuses. It exists in the catalog to prove that governance stops unsafe actions on two independent layers - access control and runtime secret scanning - rather than trusting the agent to behave.
Platform administration agent for user management, analytics, log debugging, service operations, database queries, job scheduling, and agent management via the systemprompt CLI.