AXIMUR Action Gate

What it does

AXIMUR Action Gate classifies proposed AI-agent actions as ALLOW, ESCALATE, or BLOCK before irreversible execution.

Why it exists

AI coding agents can write code, run terminal commands, publish packages, deploy infrastructure, edit payment settings, and post publicly. Those actions need a deterministic gate before execution.

Install

npm install -D @aximur/action-gate

Quick start

npx @aximur/action-gate examples/stripe_product_update.json --json

Verdicts

• ALLOW: safe local action.
• ESCALATE: human review or final click required.
• BLOCK: action is too dangerous to run automatically.

Example actions it gates

• npm publish
• Stripe product update
• AWS S3 upload
• CloudFront invalidation
• GitHub push
• public post
• email send
• package install
• payment/payout changes
• domain/DNS changes

Relationship to Package Guard

Package Guard validates package names before install. Action Gate classifies broader agent actions before execution. Package installs can be routed through Package Guard as evidence.

Package Guard: https://www.npmjs.com/package/@aximur/package-guard

Paid setup

Need help wiring Action Gate or Package Guard into a repo, CI workflow, Claude Code, Cursor, Copilot, or MCP-compatible agent setup?

AXIMUR Agent Security Setup is a one-time $499 setup service: https://buy.stripe.com/aFa7sNd9meNiadn3UIfIs05

Team Starter for hosted package-validation checks: https://buy.stripe.com/eVq3cxd9m0Ws5X70IwfIs04

Limitations

• Action Gate does not execute actions by itself.
• It does not log into accounts.
• It does not guarantee security.
• It does not replace IAM, code review, dependency scanning, or human approval.
• It is an early local policy gate for AI-agent workflows.