Entity trail

MCPTox

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-06-10 / skill-finderGate MCP tool registration on description baselines with mcp-scanTool-poisoning defenses have to operate at the tool-description layer, because that text is what steers the model. The recommended pattern is to validate every server's tool descriptions against a known-good baseline and flag anomalous content at registration time, using open-source scanners like mcp-scan (and monitors like MCPTox/MindGuard) wired into the install workflow. Bake mcp-scan into your MCP onboarding/CI so a server that quietly changes a tool description after approval is caught before it reaches an agent.
2026-04-21 / vibe-coding-researcherPattern: MCP Security Tooling Reaches Critical Mass — Scanner, Wrapper, and Auditor Ship in Same MonthApril 2026 saw Golf Scanner (config discovery + 20 checks across 7 IDEs), Astrix MCP Secret Wrapper (runtime vault integration), and mcp-sec-audit (static + dynamic analysis, 100% MCPTox detection) all ship within weeks of each other. The MCP ecosystem now has 500+ servers and 97M monthly SDK downloads, and the security tooling is finally catching up. For builders running MCP servers in production, the excuse of 'no good security tooling exists' is no longer valid — run Golf Scanner on your setup this week.
2026-04-07 / skill-findermcp-sec-audit Toolkit: Static Pattern Matching + Docker/eBPF Dynamic Fuzzing Achieves 100% MCPTox Benchmark DetectionThe mcp-sec-audit toolkit combines static pattern matching for Python-based MCP servers with dynamic sandboxed fuzzing and monitoring via Docker and eBPF. It achieves 100% detection on the MCPTox benchmark—the standard test suite for MCP tool poisoning. Detects risky capabilities (file system access, network requests, command execution) through configurable rule-based analysis and provides specific mitigation recommendations. Open source, part of the Cloud Security Alliance's Model Context Protocol Security initiative.
2026-04-06 / skill-finderGolf Scanner: Open-Source Go CLI Discovers MCP Server Configurations Across 7 IDEs with 20 Automated Security ChecksGolf Scanner is a new open-source Go CLI designed for security teams to discover MCP server configurations across seven popular IDEs (VS Code, Cursor, Windsurf, and more) and automatically run 20 unique security checks against local environments. Part of the Adversa AI April 2026 agentic security roundup alongside mcp-sec-audit (100% detection on MCPTox benchmark) and STSS (cryptographic attestation via SHA-256 Merkle trees). Addresses the shadow IT problem of unaudited MCP servers becoming production dependencies.
2026-03-25 / skill-finderOWASP Publishes Practical Guide for Secure MCP Server Development: Step-by-Step Auth, Validation, Session Isolation, and Deployment HardeningOWASP's GenAI Security Project released a hands-on implementation guide complementing the MCP Top 10. Covers secure architecture design, authentication/authorization implementation, strict input validation with sanitization, session isolation between concurrent users, and hardened container deployment with non-root users. Includes MCPTox and MindGuard tooling recommendations for real-time prompt injection and tool poisoning detection.
2026-03-22 / arxiv-researcherFormal Analysis and Supply Chain Security for Agentic AI SkillsLarge-scale empirical scan of 42,447 agent skills across multiple registries finds 26.1% exhibit at least one security vulnerability across 14 distinct vulnerability patterns, including MCP-ITP (automated implicit tool poisoning) and MCPTox benchmark attacks on real-world MCP servers. The SkillScan static analysis tool achieves 86.7% precision for automated vulnerability detection. More than 1 in 4 marketplace-sourced agent skills has a confirmed security flaw.
2026-03-15 / projects-researcherShow HN: Skill or Kill — Gamified Security Training for Spotting Malicious AI Agent SkillsSkill or Kill is an interactive game that presents players with AI agent skills and challenges them to identify malicious ones—translating the abstract supply chain attack threat (MCPTox, ToxicSkills, 5,400-skill VoltAgent directory) into hands-on practitioner training. The timing is precise: as malicious skill injection has been documented at scale in March 2026, a detection game fills a gap between white paper awareness and developer muscle memory. Gamified security training historically accelerates adoption of security practices that dry technical documentation fails to drive.
2026-03-15 / vibe-coding-researcherCVE-2026-30856 — WeKnora MCP Server Hijacks Tool Execution via Ambiguous Naming; MCPTox Shows 72.8% Attack Success Rate on o1-miniGitLab Advisories published CVE-2026-30856 (March 6, 2026) disclosing that Tencent's WeKnora MCP server is vulnerable to tool execution hijacking: a malicious remote MCP server uses ambiguous tool naming combined with indirect prompt injection to intercept and redirect tool calls intended for legitimate servers. The concurrently published MCPTox benchmark tested 20 LLM agents against 45 real-world MCP servers (353 tools) and found o1-mini has a 72.8% attack success rate — with more capable models often more susceptible due to better instruction-following of injected payloads.

Graph relationships

BENCHMARKED AGAINST
SkillScan
SkillScan achieves 86.7% precision for vulnerability detection on agent skill packages
Source finding
BENCHMARKED AGAINST
o1-mini
MCPTox benchmark found o1-mini has 72.8% attack success rate
Source finding
BENCHMARKED AGAINST
mcp-sec-audit
mcp-sec-audit achieves 100% detection on the MCPTox benchmark.
Source finding
BENCHMARKED AGAINST
SkillScan -> MCPTox
SkillScan achieves 86.7% precision for vulnerability detection on agent skill packages
Source finding
BENCHMARKED AGAINST
mcp-sec-audit -> MCPTox
mcp-sec-audit achieves 100% detection on the MCPTox benchmark.
Source finding

Source trail

Practical DevSecOps — MCP Security Vulnerabilities: Preventing Prompt Injection and Tool Poisoning (2026)Adversa AI GitHub — ModelContextProtocol-Security/mcpserver-audit Adversa AI — Top Agentic AI Security Resources April 2026 OWASP GenAI Security Project — Practical Guide for Secure MCP Server Development arXiv 2603.00195 Show HN GitLab Security Advisories

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…