# Ai Safety

> AI safety and responsible AI practices

- **Type:** Skill
- **Install:** `agentstack add skill-ssrjkk-claude-skills-ai-safety`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [ssrjkk](https://agentstack.voostack.com/s/ssrjkk)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [ssrjkk](https://github.com/ssrjkk)
- **Source:** https://github.com/ssrjkk/claude-skills/tree/main/.claude/skills/ai/ai-safety
- **Website:** https://claude.ai

## Install

```sh
agentstack add skill-ssrjkk-claude-skills-ai-safety
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# AI Safety

> Implement responsible AI practices including guardrails, monitoring, and ethical guidelines.

## Quick Start
```python
from guardrails import Guard
from guardrails.validators import Validator

class NoPIIValidator(Validator):
    def validate(self, value: str, metadata: dict) -> dict:
        import re
        # Check for emails, SSNs, credit cards
        patterns = {
            "email": r'\b[\w.+-]+@[\w-]+\.[\w.]+\b',
            "ssn": r'\b\d{3}-\d{2}-\d{4}\b',
            "credit_card": r'\b\d{4}[- ]?\d{4}[- ]?\d{4}[- ]?\d{4}\b'
        }
        found = {name: re.findall(pat, value)
                 for name, pat in patterns.items()
                 if re.search(pat, value)}
        if found:
            return {"valid": False, "error": f"PII detected: {found}"}
        return {"valid": True}

# Content moderation guard
content_guard = Guard().use(NoPIIValidator())

# Usage
result = content_guard.validate("My email is user@example.com")
print(result.error)  # "PII detected: {'email': ['user@example.com']}"
```

## Key Concepts
AI safety spans: prompt injection prevention, PII/redaction, content moderation, output validation, rate limiting, audit logging, and bias monitoring. Defense in depth — multiple layers of protection.

## When to Use
- Any production LLM deployment
- Applications handling user data or PII
- Systems where AI outputs affect real-world decisions
- Regulated industries (healthcare, finance, legal)

## Validation
1. Prompt injection attempts are blocked or sanitized
2. PII is detected and redacted in both inputs and outputs
3. Audit logs capture all LLM interactions for review
4. Rate limits prevent abuse and cost overruns

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [ssrjkk](https://github.com/ssrjkk)
- **Source:** [ssrjkk/claude-skills](https://github.com/ssrjkk/claude-skills)
- **License:** MIT
- **Homepage:** https://claude.ai

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-ssrjkk-claude-skills-ai-safety
- Seller: https://agentstack.voostack.com/s/ssrjkk
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.