Read-only safety inspection of a third-party MCP server BEFORE it is added to a config or trusted. Reads the server's declared tool descriptions, input schemas, scopes, transport, and env/secret surface (not just its README), compares declared behavior against what the tools actually expose, scans for tool-description poisoning and prompt injection, over-broad or undeclared scopes, egress and credential access, config tampering, and hidden Unicode, and returns an add/decline/sandbox verdict for a human to approve. Never installs, never auto-trusts. Use when evaluating an external MCP server before adding it to .mcp.json or enabling it. Do not use to vet a third-party agent skill or plugin (use skill-vet), to review first-party product code (use review-hard or security-review), or to fetch a remote server definition from the web (record its origin via capture-references, then obtain a local copy out of band into a quarantine directory first).
Install with
Claude CodeCodexCursorGitHub CopilotGemini CLI
Supported by the Atlas install prompt. Skill-specific requirements vary.
Skill stats
Skills1CategorySecurityStars0
About this skill
Read-only safety inspection of a third-party MCP server BEFORE it is added to a config or trusted. Reads the server's declared tool descriptions, input schemas, scopes, transport, and env/secret surface (not just its README), compares declared behavior against what the tools actually expose, scans for tool-description poisoning and prompt injection, over-broad or undeclared scopes, egress and credential access, config tampering, and hidden Unicode, and returns an add/decline/sandbox verdict for a human to approve. Never installs, never auto-trusts. Use when evaluating an external MCP server before adding it to .mcp.json or enabling it. Do not use to vet a third-party agent skill or plugin (use skill-vet), to review first-party product code (use review-hard or security-review), or to fetch a remote server definition from the web (record its origin via capture-references, then obtain a local copy out of band into a quarantine directory first).
Quick start
Copy the install prompt. Use the green button above this README.
Send it to your agent. Paste it into your coding agent’s chat and send.
Review the setup. Your agent checks the skill’s instructions, requirements, and installation scope.
Read-only safety inspection of a third-party MCP server BEFORE it is added to a config or trusted. Reads the server's declared tool descriptions, input schemas, scopes, transport, and env/secret surface (not just its README), compares declared behavior against what the tools actually expose, scans for tool-description poisoning and prompt injection, over-broad or undeclared scopes, egress and credential access, config tampering, and hidden Unicode, and returns an add/decline/sandbox verdict for a human to approve. Never installs, never auto-trusts. Use when evaluating an external MCP server before adding it to .mcp.json or enabling it. Do not use to vet a third-party agent skill or plugin (use skill-vet), to review first-party product code (use review-hard or security-review), or to fetch a remote server definition from the web (record its origin via capture-references, then obtain a local copy out of band into a quarantine directory first).
About this listing
Category
Security
Reported license
MIT
Listing language (estimated)
English
Review status
Not individually reviewed
Agent compatibility
Atlas installation supports Claude Code, Codex, and Cursor. Skill-specific dependencies and features are checked during setup.
Claude Code
Codex
Cursor
GitHub Copilot
Gemini CLI