Metadata-Version: 2.4
Name: sprk3-defend-agents
Version: 0.1.0
Summary: Runtime security for AI agents — detects jailbreak, prompt injection, and tool misuse from inside the agent.
Project-URL: Homepage, https://defend.sprk3.com/agents
Project-URL: Documentation, https://docs.sprk3.com/defend-agents
Project-URL: Repository, https://github.com/sprk3/defend-agents
Author-email: SPR{K}3 Security Research <support@sprk3.com>
License-Expression: MIT
Keywords: agents,ai,anthropic,langchain,openai,prompt-injection,security
Requires-Python: >=3.9
Requires-Dist: httpx>=0.25.0
Provides-Extra: all
Requires-Dist: anthropic>=0.30.0; extra == 'all'
Requires-Dist: langchain-core>=0.2.0; extra == 'all'
Requires-Dist: openai>=1.30.0; extra == 'all'
Provides-Extra: anthropic
Requires-Dist: anthropic>=0.30.0; extra == 'anthropic'
Provides-Extra: langchain
Requires-Dist: langchain-core>=0.2.0; extra == 'langchain'
Provides-Extra: openai
Requires-Dist: openai>=1.30.0; extra == 'openai'
Description-Content-Type: text/markdown

# SPR{K}3 Defend Agents

Runtime security for AI agents — detects jailbreak, prompt injection, and tool misuse from inside the agent, not at the gateway.

## Install

```bash
pip install sprk3-defend-agents
```

## Quick Start

```python
from sprk3.defend import monitor
monitor(api_key="sk-...")
```

That's it. Auto-instruments Anthropic, OpenAI, and LangChain clients.

## What it detects

- **Indirect prompt injection** — hidden instructions in web pages, emails, docs, tool outputs
- **Tool misuse** — legitimate tools used on attacker's behalf
- **Agent tamper** — heartbeat detects monitoring disabled or bypassed
- **Behavioral anomaly** — tool velocity spikes, entropy shifts, new tool usage

## Privacy

All detection runs client-side. Server sees metadata only. Evidence stays on your machine.

## Links

- [defend.sprk3.com](https://defend.sprk3.com/agents)
- [support@sprk3.com](mailto:support@sprk3.com)
