Skills

Teach your agent to use Puffer.

The Puffer skill gives AI agents the knowledge to self-install Puffer, interpret defense layer verdicts, and respect operating mode boundaries.

SKILL.md (Claude Code) puffer-skill.md (Universal) puffer-skill.json (Structured)

Integration Guide

Pick your agent and follow the steps.

$ mkdir -p ~/.claude/skills/puffer

$ curl -fsSL https://openpuffer.org/skill/SKILL.md -o ~/.claude/skills/puffer/SKILL.md

Claude Code auto-discovers skills in ~/.claude/skills/. No further config required.

SKILL.md Preview

What your agent reads after install.

Download SKILL.md

SKILL.md

---
name: puffer
description: Self-install and use Puffer (the autonomous immune system for AI agents)
  to monitor your own actions for PII leaks, prompt injection, dangerous commands,
  network egress, filesystem violations, runaway behavior, and unauthorized MCP servers.
allowed-tools: Bash, Read, Write
---

# Puffer Skill — Agent Self-Onboarding

## Step 1 — Detect existing install
```bash
which puffer || echo "missing"
```

## Step 2 — Install if missing
```bash
curl -fsSL https://openpuffer.org/install.sh | bash
```

## Step 3 — Initialize in monitor mode
```bash
puffer init
```
CRITICAL: Puffer always defaults to monitor mode.
NEVER auto-run `puffer mode enforce` without explicit user permission.

## Step 4 — Verify
```bash
puffer status
```

## Step 5 — Stream audit log
```bash
puffer logs -f &
```

# ... (see full SKILL.md for verdict tables, guardrails, and mode docs)

Back to openpuffer.org