# OPFOR

> OPFOR is an open-source adversary simulation framework — a dedicated opposing force that runs continuous multi-turn attack campaigns against your AI agents, 24/7.

OPFOR probes agents with prompt injection, jailbreaks, system prompt extraction, tool-call hijacking, PII leakage, policy circumvention, and social-engineering pivots across turns. Every successful exploit is severity-graded, mapped to compliance frameworks, and auto-converted into a permanent regression test.

Built by KeyValue Software Systems.

---

## Marketing Site (https://agentopfor.ai)

Authoritative product pages:

- [Homepage](https://agentopfor.ai/): Open-source adversary simulation for AI agents
- [Documentation](https://docs.agentopfor.ai/): Setup, configuration, and integration guides
- [GitHub](https://github.com/KeyValueSoftwareSystems/opfor): Source code and contributions
- [Try OPFOR](https://github.com/KeyValueSoftwareSystems/opfor): Get started with adversarial testing
- [Talk to us / Demo](https://discord.com/invite/Sb8xfmrjjW): Sales and demo requests
- [Pricing](https://www.getnetra.ai/pricing/): Plans and enterprise options
- [About](https://www.getnetra.ai/about-us/): Company information
- [Contact](https://www.getnetra.ai/contact-us/): Contact and support
- [Privacy Policy](https://www.getnetra.ai/privacy-policy/)

AI discovery files:
- [llms.txt](https://agentopfor.ai/llms.txt) — this file
- [llm.txt](https://agentopfor.ai/llm.txt) — alias index
- [ai.txt](https://agentopfor.ai/ai.txt) — AI crawler manifest
- [Sitemap](https://agentopfor.ai/sitemap.xml)

---

## Product Capabilities

1. **Continuously Evolving Attack Library** — Tracks OWASP LLM Top 10, latest jailbreak research, and public exploit databases.
2. **Multi-Turn Adversarial Agents** — Surfaces top recurring breakage patterns across millions of sessions.
3. **Full Attack Coverage** — Flags sessions with anomalous token usage or quality degradation automatically.
4. **Severity-Graded Findings** — Spots deviations in tool usage, latency, or quality the moment they emerge.
5. **Auto-Generated Regression Tests** — Every successful exploit becomes a permanent test case in your pipeline.
6. **Continuous, Not One-Time** — Runs every deploy and every night; pentests go stale in a week.
7. **Compliance-Ready Reports** — Maps findings to OWASP, NIST AI RMF, and EU AI Act categories.
8. **Patch Verification Loop** — Every fix is re-attacked automatically until the security barrier holds.

---

## Attack Coverage

- Direct and indirect prompt injection
- Jailbreaks and system prompt extraction
- Tool-call hijacking and misuse
- PII and secret leakage
- Policy circumvention
- Multi-turn social engineering pivots

---

## Compliance Mapping

- OWASP LLM Top 10
- NIST AI RMF
- EU AI Act categories

---

## Contact

- Sales: contact@agentopfor.ai
- Security: security@getnetra.ai
- Documentation: https://docs.agentopfor.ai/
- GitHub: https://github.com/KeyValueSoftwareSystems/opfor

---

## Social

- LinkedIn: https://www.linkedin.com/company/getnetra/
- X (Twitter): https://x.com/getnetraAI
- YouTube: https://www.youtube.com/@getnetra

---

## Parent Organization

- KeyValue Software Systems: https://www.keyvalue.systems/