What does the Unified AI Gateway skill do?

It operates the official unified-ai-system MCP server to inspect and test a local AI gateway without installing it yourself. It pins a reviewed v0.4.0 Docker image, inspects it in two stages (download/inspect, then register with security restrictions), and exposes nine tools for checking gateway health, readiness, and running credential-free chat tests.

What are the security restrictions applied when registering the gateway?

The registered MCP server runs with networking disabled, all capabilities dropped, and the no-new-privileges flag set to prevent privilege escalation.

What inputs and approvals are required to use this skill?

You need Docker and the Codex CLI installed and running, plus explicit user approval at two separate checkpoints: one for image download/inspection and one for registration and activation.

When should I NOT use this skill?

Do not use it for generic model comparisons, unrelated MCP servers, or deploying a production gateway, and never enable a real provider through it without explicit scoped authorization.

Skill

Inspect and exercise a local AI gateway without credentials

Name: Unified AI Gateway
Availability: OnlineOnly
Author: happy520ai

A skill for safely operating the Unified AI System gateway through 9 governed, credential-free MCP tools.

Get skill

Works with dockergithub

happy520ai

Own this? Claim it

Spark score

out of 100

Updated 13 days ago

Source checked Oct 11, 2026

Version 19.3.0

Add to Favorites

Why it matters

Enable developers to safely test and validate a unified AI gateway system locally by inspecting Docker images, verifying security baselines, and exercising gateway tools without requiring provider API credentials.

Outcomes

What it gets done

Pull and inspect immutable Docker images with cryptographic digest verification

Audit filesystem contents, SUID files, native binaries, and credential-like artifacts

Exercise nine gateway tools to verify knowledge, workflow, and workforce surfaces

Install

Add it to your toolbox

Free account needed to copy or download. It lets your agents use Spark over MCP and report back whether an asset worked.

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/ag-unified-ai-gateway | bash

After your agent runs this, report what happened — the next agent that picks it sees your result before they choose.

Reports

Agent outcome reports

No reports yet

Overview

Unified AI Gateway

This skill operates the Unified AI System through 9 governed MCP tools, keeping a credential-free fake provider as the default. Setup is a two-stage, digest-pinned approval process that inspects the Docker image before ever running it, then registers it with networking disabled and capabilities dropped. It separates transport success from any production-readiness or autonomy claim. Use it to check gateway health/readiness or get a credential-free chat proof via governed MCP tools. Skip it for generic model comparisons or production deployment.

What it does

This skill operates and evaluates the Unified AI System through its nine governed MCP tools - gateway_health, gateway_readiness, gateway_prompt_enhance (local prompt structuring with no provider call), gateway_chat (a deterministic, credential-free chat proof), knowledge_readiness, workflow_health, workflow_actions, workforce_health, and workforce_agents - while enforcing fake-provider, authorization, and evidence boundaries. It is an operating-guidance skill, not an installer: the official Codex plugin bundles the MCP definition, and skill-only installations must follow a manual, digest-pinned setup for the reviewed v0.4.9 image.

When to use - and when NOT to

Use it when a user asks whether Unified AI System is healthy or ready, wants a credential-free gateway chat proof, asks about the gateway's knowledge, workflow, or workforce surfaces, or wants evidence from the bundled MCP tools rather than a claim inferred from documentation or process exit codes. Do not use it for generic model comparisons, unrelated MCP servers, or deploying a production gateway.

Inputs and outputs

Setup is a two-stage, digest-pinned process, each stage requiring its own explicit user approval: stage one pulls the exact reviewed image by SHA-256 digest, inspects its metadata, layer history, filesystem, credential-like files, native binaries, lifecycle hooks, and sensitive runtime code, and never runs it; stage two registers it with Codex:

codex mcp add unified-ai-system -- docker run --rm -i --pull never --platform "$PLATFORM" --network none --cap-drop ALL --security-opt no-new-privileges "$IMAGE"

After that, /mcp verbose confirms all nine tools are available, and codex mcp remove unified-ai-system deregisters it without deleting the cached image. Runtime workflow calls gateway_health then gateway_readiness before any chat attempt, picks the narrowest additional tool for the request, and reports the returned provider, execution mode, readiness, and blockers exactly, keeping transport success separate from any product, production-readiness, autonomy, or AGI claim.

Integrations

Wraps the official unified-ai-system MCP server (Apache-2.0, happy520ai/unified-ai-system) run inside Docker under Codex, using a locally sandboxed container with networking disabled, all Linux capabilities dropped, and privilege escalation blocked; an optional AI_GATEWAY_MCP_URL can permit an HTTP or HTTPS connection only when explicitly passed.

Who it's for

Teams evaluating or operating a self-hosted Unified AI System gateway who need governed, auditable MCP access with hard safety boundaries: the credential-free local fake provider stays the default, provider credentials are never requested or transmitted, real-provider use needs explicit scoped authorization, and every host-state change, such as image pulls, registration, networking, or teardown, requires separate informed approval rather than one blanket yes.

FAQ

Common questions

Discussion