nio

Scan codebases and evaluate runtime actions for execution risks.

19|Updated Mar 30, 2026
One-click install
npx skills add https://github.com/core0-io/nio --skill nio
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: nio
Source: https://github.com/core0-io/nio/tree/main/plugins/claude-code/skills/nio
Command: npx skills add https://github.com/core0-io/nio --skill nio

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Nio helps you prevent autonomous AI agents from executing risky tool calls by evaluating them before they run and recording auditable decisions you can review later.

Core Features & Use Cases

  • Action safety evaluation (Phase 0–6): Checks exec commands, network requests, secrets, and file operations and returns allow/deny/confirm decisions based on configurable policy.
  • Code scanning for execution risks: Scans a codebase for known patterns such as shell execution, prompt injection, credential leakage, webhooks/exfiltration domains, and obfuscation.
  • Observability and audit logging (OTEL + JSONL): Captures tool calls as OpenTelemetry metrics/traces and writes a local audit trail for post-incident review.
  • Use Case: When deploying an agent to production with Claude Code, Codex CLI, OpenClaw, or Hermes, you can proactively block dangerous actions and surface clear evidence for every decision.

Quick Start

Run: nio scan ./your-project to produce an execution risk report for the codebase.

Frequently Asked Questions about nio

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent autonomous AI agents from executing risky tool calls?

Preventing risky tool calls requires evaluating agent actions before execution using a Phase 0–6 decision pipeline that scans exec commands, network requests, secrets, and file operations to return allow, deny, or confirm decisions based on configurable policy.

How do I scan a codebase for prompt injection and credential leakage patterns?

Scanning a codebase for prompt injection and credential leakage involves running a pre-deployment risk audit that detects known shell execution, exfiltration domains, and obfuscation patterns to produce a detailed execution risk report.

Does Nio work with Claude Code, Codex CLI, and OpenClaw for agent safety?

Yes, agent safety evaluation applies to tool-call interception across common platforms including Claude Code, Codex CLI, OpenClaw, and Hermes, proactively blocking dangerous actions during production deployment.

Can I export OpenTelemetry metrics and traces for AI agent audit logging?

Yes, you can export OpenTelemetry metrics and traces for audit logging by capturing tool calls as OTEL data and writing a local JSONL audit trail to provide clear evidence for post-incident review of every decision.

What is execution assurance for autonomous AI agents?

Execution assurance is the process of evaluating runtime actions for safety before they execute, applying guard policies to intercept tool calls and producing detailed findings, audit events, and OpenTelemetry observability data.

What are the limitations of applying guard policies to agent tool-call interception?

Limitations of guard policies include relying on configurable policy rules for allow, deny, or confirm decisions, meaning the Phase 0–6 pipeline evaluates known patterns like shell execution and exfiltration domains but may require custom rules for novel injection vectors.