AI-Infra-Guard (A.I.G) — Tencent Zhuque Lab

What it is

Deployment and operational specifics for AI-Infra-Guard that the feature summary omits: how to stand it up, how to point it at a target, and one hard security constraint. A.I.G runs as a Docker-deployed web UI on port 8088, plus pip-installable CLIs and a Go-built unified CLI.

What to do

Three install paths. Pre-built images (fastest):

git clone https://github.com/Tencent/AI-Infra-Guard.git
cd AI-Infra-Guard
docker-compose -f docker-compose.images.yml up -d

One-click script (auto-installs Docker):

curl https://raw.githubusercontent.com/Tencent/AI-Infra-Guard/refs/heads/main/docker.sh | bash

Skill-scan as a standalone package:

pip install aig-skill-scan
export LLM_API_KEY="your-api-key"
aig-skill-scan --repo /path/to/your/skill -m deepseek-v4-flash --language en -o result.json

The web UI then serves at http://localhost:8088; API docs at http://localhost:8088/docs/index.html, API-Checker docs at http://127.0.0.1:8088/api-checker/docs.

Usage gotcha the page flags explicitly: for infra scanning, the target is the network address of a running AI service (e.g. http://127.0.0.1:8000 for a live vLLM instance), not a GitHub URL or source path — A.I.G connects to the live service and fingerprints its version against the CVE set. MCP scanning is the exception: it accepts a remote repo URL or a local archive with no running instance.

Skill-scan vulnerability output maps to the SkillTrustBench T01–T09 taxonomy (9 skill-risk categories), which aligns it for CI/CD gating.

Caveats

The page states a deployment constraint the summary does not: A.I.G currently lacks any authentication mechanism and should not be deployed on public networks — it is positioned for internal enterprise/individual use only. Treat any exposed instance as open.

Apache 2.0, but with attribution strings required: integrating the scanning engine into a product obliges the notice "This project integrates AI-Infra-Guard, open-sourced by Tencent Zhuque Lab.", and research/report use must name "Tencent Zhuque Lab AI-Infra-Guard" with a repo link. A "Pro" tier (aigsec.ai) is gated behind invitation codes for contributors.