# Pix quickstart

Wire the Pix hooks into Claude Code or Codex, install the 18 skills, check both are live, and run your first skill in the agent.

This page puts Pix into the coding agent you already use: the hooks that stop a malicious install or a committed credential, and the skills you call by name.

## Prerequisites

- The Vulnesis CLI on the `PATH` of the process that starts your agent: `brew install Vulnesis/tap/vulnetix`, or see [Install the CLI](/cli/get-started/install.md).
- A coding agent on this machine. Hooks are wired for Claude Code and OpenAI Codex. Skills reach every agent in [What each agent gets](/pix/get-started/install.md#what-each-agent-gets).
- No credential. Without one the CLI uses the shared Community tier.

## 1. Wire the hooks

```bash
vulnetix agent install
```

It detects each agent and writes one `vulnetix agent hook` entry per event into that agent's own settings file:

```text
Claude Code
  hooks    configured ~/.claude/settings.json
  skills   ~/.claude/skills
  mcp      https://mcp.vulnes.is/mcp
```

A second run prints `hooks    already configured`. Add `--dry-run` to see the change first.

## 2. Install the skills

```bash
vulnetix skills install
```

`vulnetix pix install` does the same. It uses `npx skills add`, else the `claude` plugin commands, else `gh skills add`, whichever it finds first. In Claude Code you can instead run `/plugin marketplace add Vulnesis/pix-ai-coding-assistant` and `/plugin install vulnetix@vulnetix-plugins`, which also brings the hooks and the 5 subagents.

## 3. Check both are live

```bash
vulnetix agent hosts
vulnetix skills check
```

`agent hosts` should show `INSTALLED yes` for your agent, and `HOOKS yes` for Claude Code and Codex. `skills check` lists the Vulnesis skills in each skills directory, and any that are missing or retired. In Claude Code, `/hooks` lists three `vulnetix agent hook` entries.

## 4. Run a skill

Open your agent in a repository and ask whether a package name is safe:

```text
/vulnetix:typosquat-check expresss
```

That is the Claude Code plugin's name. Installed through `npx skills` or `gh`, the skill is `/typosquat-check`. The skill answers `ALLOW`, `WARN` or `BLOCK` with its reason, and records the check in `.vulnetix/memory.yaml`.

## 5. Try the guard

Ask the agent to run `npm i event-stream@3.3.6`. That version is on a malware feed, so the dependency guard denies the call and gives the agent the package card as the reason. `npm i axios` passes and prints nothing: silence is a pass.

## Next

- [Install, update and verify](/pix/get-started/install.md) for every agent, updates and pruning.
- [Skills](/pix/skills/overview.md) for all 18 and their arguments.
- [Dependency guard](/pix/hooks/dependency-guard.md) and the [policy file](/pix/hooks/policy-file.md) to warn instead of block.
- [Troubleshooting](/pix/get-started/troubleshooting.md) when a hook says nothing.
