Registry -- AgentSign

Framework	Category	Stars	Verdict	Risk	OWASP	Findings	Scanned
AutoGPT	Autonomous Agent	182K+	✗ FAIL	65	AA-03 AA-09	exec(), os.system(), no sandbox	2026-03-10
BabyAGI	Autonomous Agent	20K+	✗ FAIL	55	AA-03 AA-06	exec(), eval(), supply chain	2026-03-10
Open Interpreter	Code Agent	57K+	✗ FAIL	80	AA-03 AA-09 AA-01	exec(), os.system(), child_process, no sandbox	2026-03-10
MetaGPT	Multi-Agent	48K+	✗ FAIL	50	AA-03 AA-09	subprocess shell=True, no sandbox	2026-03-10
GPT Engineer	Code Agent	53K+	✗ FAIL	55	AA-03 AA-06	exec(), subprocess, supply chain	2026-03-10
ChatDev	Multi-Agent	27K+	✗ FAIL	50	AA-03 AA-09	exec(), subprocess	2026-03-10
SuperAGI	Autonomous Agent	15K+	✗ FAIL	50	AA-03 AA-01	subprocess shell=True, excessive agency	2026-03-10
CAMEL	Multi-Agent	6K+	✗ FAIL	40	AA-03 AA-06	eval(), supply chain	2026-03-10
AgentGPT	Autonomous Agent	32K+	✗ FAIL	55	AA-03 AA-09 AA-06	eval(), no sandbox, supply chain	2026-03-10
LangChain	Agent Framework	100K+	~ WARN	26	AA-06 AA-02	Supply chain, prompt injection vectors	2026-03-10
CrewAI	Multi-Agent	45K+	~ WARN	23	AA-06 AA-04	Supply chain, excessive permissions	2026-03-10
Microsoft AutoGen	Multi-Agent	50K+	~ WARN	23	AA-03 AA-06	Code execution patterns, supply chain	2026-03-10
Semantic Kernel	Agent Framework	23K+	~ WARN	18	AA-06 AA-07	Supply chain, output handling	2026-03-10
TaskWeaver	Code Agent	5K+	~ WARN	23	AA-03 AA-09	Eval patterns, no sandbox	2026-03-10
OpenAI Agents SDK	Agent Framework	19K+	~ WARN	11	AA-06 AA-01	Supply chain, excessive agency config	2026-03-10
Haystack	Agent Framework	18K+	~ WARN	11	AA-06 AA-07	Supply chain, output handling	2026-03-10
Smolagents	Agent Framework	15K+	~ WARN	18	AA-03 AA-06	Code execution, supply chain	2026-03-10
Swarm	Multi-Agent	18K+	~ WARN	11	AA-01 AA-06	Excessive agency, supply chain	2026-03-10
Google ADK	Agent Framework	15K+	✓ PASS	3	AA-06	Minor supply chain note	2026-03-10
Anthropic Claude SDK	SDK	8K+	✓ PASS	0	--	No issues detected	2026-03-10
Vercel AI SDK	SDK	12K+	✓ PASS	0	--	No issues detected	2026-03-10
LlamaIndex	Agent Framework	38K+	✓ PASS	3	AA-06	Minor supply chain note	2026-03-10
Phidata	Agent Framework	18K+	✓ PASS	0	--	No issues detected	2026-03-10
Instructor	SDK	9K+	✓ PASS	0	--	No issues detected	2026-03-10
Mastra	Agent Framework	10K+	✓ PASS	0	--	No issues detected	2026-03-10
Amazon Bedrock Agents	Cloud Agent	N/A	✓ PASS	3	AA-06	Minor supply chain note	2026-03-10
AgentSign SDK	Identity SDK	OSS	✓ PASS	0	--	No issues detected	2026-03-10

Methodology & Legal

About this registry

The AgentSign Registry performs automated static analysis of publicly available open-source code. All scanned projects are sourced from public GitHub repositories under their respective open-source licences. This is the same approach used by established security scanning services such as Snyk, Sonatype OSS Index, and the GitHub Advisory Database.

What we scan for: Code patterns mapped to the OWASP Agentic AI Top 10 (2025) and OWASP MCP Top 10. These are industry-standard security benchmarks, not proprietary criteria.

Verdicts reflect automated pattern matching at a point in time and may not reflect the current state of any project
A FAIL verdict indicates the presence of code patterns flagged by OWASP guidelines -- it does not mean the software is insecure, exploitable, or defective
Many flagged patterns (e.g. exec(), subprocess) are intentional design decisions in agent frameworks that execute code by nature
This is not a penetration test, runtime assessment, or security audit
No proprietary, private, or non-public code is accessed or scanned
All findings are based on publicly available source code

Fair use and responsible disclosure: This registry is published in good faith for the purpose of improving agentic AI security. If you are a maintainer of a listed project and believe a finding is inaccurate or wish to discuss remediation, please contact contact@agentsign.dev. We will promptly review and update any disputed findings.

AgentSign Registry is provided "as is" without warranty. CyberSecAI Ltd accepts no liability for decisions made based on registry data. All trademarks and project names are the property of their respective owners.

Get Started

Scan your agent in 60 seconds

No signup. No API key. Just add one file to your repo.

Add the GitHub Action

Create .github/workflows/agentsign.yml in your repo:

# .github/workflows/agentsign.yml
name: AgentSign Security Scan
on: [push, pull_request]

jobs:
  scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: razashariff/agentsign-action@v1
        with:
          path: '.'
          fail-on: 'FAIL'

Every push and PR gets scanned against 12 OWASP rules. FAIL blocks the merge.

Add a badge to your README

Show your security status. Copy this line:

![AgentSign](https://img.shields.io/endpoint?url=https://registry.agentsign.dev/api/badge/YOUR-AGENT-NAME)

Replace YOUR-AGENT-NAME with your agent name. PASS = green, WARN = yellow, FAIL = red.

Or scan via API

POST your code directly. No auth needed:

curl -X POST https://registry.agentsign.dev/api/scan \
  -H "Content-Type: application/json" \
  -d '{"code": "YOUR_CODE_HERE", "name": "my-agent"}'

Returns verdict, risk score, and findings mapped to OWASP Agentic AI Top 10.

View on GitHub API Reference

API

Scan any agent in seconds

Public endpoints. No auth required. Rate-limited.

GET /api/badge/:name

shields.io-compatible badge. PASS = green, WARN = yellow, FAIL = red. Cached 5 minutes.

curl https://registry.agentsign.dev/api/badge/autogpt

{ "schemaVersion": 1, "label": "AgentSign", "message": "FAIL", "color": "red" }

POST /api/scan

Scan agent code against 12 OWASP rules. Max 50KB.

curl -X POST https://registry.agentsign.dev/api/scan \
  -H "Content-Type: application/json" \
  -d '{"code": "exec(user_input)", "name": "my-agent"}'

{ "verdict": "FAIL", "risk_score": 40, "findings": [...] }

GET /api/rules/version

Current rules version and count.

{ "version": "1.0.0", "rule_count": 12 }

GET /api/registry

Full registry as JSON. All 27 frameworks with verdicts and scores.

9 of 27 agent frameworks fail security scan