ProofRange
SaaS-style financial ops app: stateful IDOR, blind SSRF, prototype pollution → RCE, JWT attacks, cache poisoning, and business-logic races. Built to stress scanners that only do unauthenticated crawl.
Open lab →VulnGym
Train on real attack chains.
Measure what actually breaks.
Deliberately vulnerable applications for manual pentesting and benchmarking autonomous security tools. Each ring adds difficulty — from API logic to blind OOB chains.
⚠ Labs include real RCE and SSRF. Use isolated environments only. Never point scanners at systems you do not own.
23planted vulns · ProofRange
10planted vulns · AgentRange
Oracletruth.json + coverage.js
Rings
AgentRange
10 vulns on agent-style surfaces: GraphQL introspection, alias IDOR, mutation chains,
MCP http_fetch SSRF,
scope HPP, and double-refund logic. Remote oracle: 10/10 verified.
MeshRange
Distributed-app mesh: gRPC metadata trust, WebSocket subscribe IDOR, OAuth code replay, SSRF to internal worker, multi-tenant header override, and admin RPC via reflection. Remote oracle: 10/10 verified.
Open lab → PENTEST.md →Start here
Manual pentest — no scanner required. Log in to ProofRange, follow the guide, track your score.
Builders: GitHub · VULNGYM.md · Scanner benchmark (optional)