# msoedov/agentic_security

Agentic LLM Vulnerability Scanner / AI red teaming kit 🧪

Repository: https://github.com/msoedov/agentic_security
Canonical: https://ross.abutalabs.com/products/agentic_security
Homepage: https://agentic-security.vercel.app
Language: Python
License: Apache-2.0
License Family: permissive
Topics: llm-guardrails, llm-security, llm-jailbreaks, llm-scanner, llm-vulnerabilities, llm-fuzzer, llm-fuzzing, llm-fuzzer-aggregator, ai-red-team, llm-evaluation, llm-evaluation-framework, prompt-testing, agent-security, agent-framework
Last push: 2026-08-18T16:15:08+00:00

## Health v2 (maintenance only)
Score: 87/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 87, longevity 62
- inputs: {"age_days": 874, "days_push": 15, "days_rel": 91, "gap_med": 12.5, "n_releases_24m": 15}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1977, forks 280 (observed 2026-08-28T04:06:01.454168+00:00)

## What it is
Agentic Security is an open-source LLM vulnerability scanner and AI red-teaming toolkit that probes large language models and agent workflows for jailbreaks, prompt injection, and multimodal attacks. It ships as a Python CLI and web UI that fuzzes any LLM API endpoint using adversarial prompt datasets, multi-step attacks, and RL-based probes.

## Use cases
- red team my llm app for jailbreaks
- scan llm api for prompt injection vulnerabilities
- fuzz a chatbot with adversarial prompts
- test llm safety guardrails before deployment
- run owasp llm top 10 security checks
- stress test llm endpoint with attack prompts
- evaluate multimodal attack robustness of my model

## When to choose
- you need automated adversarial testing of an LLM API or agent workflow
- you want a self-hosted scanner with a web UI and customizable prompt datasets
- you're doing AI red-teaming or OWASP LLM Top 10 assessments

## When to avoid
- you need runtime protection or guardrails in production rather than pre-deployment scanning
- you want a guaranteed defense - the tool itself warns it is a scanner, not a foolproof solution
- you need static code analysis of your application rather than black-box LLM probing

## Facets
- artifact type: cli-tool
- maturity: active
- function: security, vulnerability-scanning, penetration-testing, fuzzing, llm-inference, testing
- domain: security, large-language-models, penetration-testing, developer-tools
- platform: python, cli, self-hosted
- tags: llm-security, red-teaming, jailbreak, prompt-injection, adversarial-attacks, llm-evaluation, owasp-llm-top-10, multimodal-attacks, ai-agents, web-server, docker

## Member repositories
- msoedov/agentic_security (main) score 87

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:01.454168+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:04:47.836950+00:00, confidence not recorded.
  - readme: https://github.com/msoedov/agentic_security (fetched 2026-08-28T04:06:01.454168+00:00, sha 4d400dc0c7fa)
  - homepage: https://agentic-security.vercel.app (fetched 2026-08-29T10:43:48.164790+00:00, sha 3cc9ce2fa68f)
  - registry_pypi: https://pypi.org/pypi/agentic_security/json (fetched 2026-08-29T10:43:48.175699+00:00, sha 8c5ba135d9b4)
- Data as of 2026-08-30T08:39:29.467469+00:00.