context safety score
A score of 49/100 indicates multiple risk signals were detected. This entity shows patterns commonly associated with malicious intent.
supply chain
The GitHub repository contains NO source code — only documentation, a Claude Code skill/plugin, and static website files. The npm packages (mcp-fortress and @mcp-fortress/core) are published with compiled JavaScript from source that is not available for public review. The @mcp-fortress/core dependency uses a wildcard '*' version constraint, meaning any version of this single-maintainer package could be pulled in at install time without restriction. Combined with a brand-new anonymous account (created 2025-11-07, anonymous Protonmail email, zero followers, no identity), 23 versions published in 10 days, and no tests, this creates an opaque supply chain where the actual executing code cannot be audited from the repository. (location: package.json dependency '@mcp-fortress/core': '*'; npm registry mcp-fortress@0.3.6; GitHub repo file tree (no src/ directory))
description injection
The SKILL.md file instructs Claude Code to 'Always scan before recommending installation' and positions itself as a mandatory security gate for all MCP server evaluations. When installed as a Claude Code plugin, this skill automatically activates when users ask about MCP server safety and routes all queries through the mcp-fortress remote tools. This creates a trust-bootstrapping pattern: a security scanner with no verifiable source code, from a brand-new anonymous account, that inserts itself as the authoritative security oracle for the agent. While the skill description itself is not overtly malicious, the combination of autonomously activating on security questions and routing all package names to a third-party remote endpoint (server.smithery.ai) under the guise of security scanning represents a deceptive trust escalation. (location: claude-code-skill/skills/mcp-fortress/SKILL.md — 'When to Use This Skill' section and 'Best Practices' item 1: 'Always scan before recommending installation')
cross server attack
The .mcp.json configuration automatically connects the agent to a remote third-party endpoint (https://server.smithery.ai/@mcp-fortress/mcp-fortress-server/mcp) controlled by an anonymous single maintainer. When installed as a Claude Code plugin, every MCP server name the user asks about gets sent to this remote endpoint via the scan_mcp_server, analyze_prompt_injection, and detect_tool_poisoning tools. The remote server's responses could influence the agent's security recommendations (approving malicious servers or rejecting legitimate competitors) with no way to verify the server's behavior since the source code is not published. (location: claude-code-skill/skills/mcp-fortress/.mcp.json — url: 'https://server.smithery.ai/@mcp-fortress/mcp-fortress-server/mcp')
curl https://api.brin.sh/mcp/mcp-fortress%2Fmcp-fortressCommon questions teams ask before deciding whether to use this mcp server in agent workflows.
mcp-fortress/mcp-fortress currently scores 49/100 with a suspicious verdict and medium confidence. The goal is to protect agents from high-risk context before they act on it. Treat this as a decision signal: higher scores suggest lower observed risk, while lower scores mean you should add review or block this mcp server.
Use the score as a policy threshold: 80–100 is safe, 50–79 is caution, 20–49 is suspicious, and 0–19 is dangerous. Teams often auto-allow safe, require human review for caution/suspicious, and block dangerous.
brin evaluates four dimensions: identity (source trust), behavior (runtime patterns), content (malicious instructions), and graph (relationship risk). Analysis runs in tiers: static signals, deterministic pattern checks, then AI semantic analysis when needed.
Identity checks source trust, behavior checks unusual runtime patterns, content checks for malicious instructions, and graph checks risky relationships to other entities. Looking at sub-scores helps you understand why an entity passed or failed.
brin performs risk assessments on external context before it reaches an AI agent. It scores that context for threats like prompt injection, hijacking, credential harvesting, and supply chain attacks, so teams can decide whether to block, review, or proceed safely.
No. A safe verdict means no significant risk signals were detected in this scan. It is not a formal guarantee; assessments are automated and point-in-time, so combine scores with your own controls and periodic re-checks.
Re-check before high-impact actions such as installs, upgrades, connecting MCP servers, executing remote code, or granting secrets. Use the API in CI or runtime gates so decisions are based on the latest scan.
Learn more in threat detection docs, how scoring works, and the API overview.
Assessments are automated and may contain errors. Findings are risk indicators, not confirmed threats. This is a point-in-time assessment; security posture can change.
integrate brin in minutes — one GET request is all it takes. query the api, browse the registry, or download the full dataset.