mcp-installer
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- REMOTE_CODE_EXECUTION (HIGH): The skill is designed to find and configure
npxcommands from untrusted GitHub repositories. Usingnpx -y [package]@latestdownloads and executes code immediately upon use. There is no verification of package integrity or author reputation, allowing an attacker to distribute malicious MCP servers. - COMMAND_EXECUTION (HIGH): The skill modifies the
~/.claude.jsonconfiguration file. This file defines tools that the agent can execute. By adding entries to themcpServerssection, the skill creates a persistent execution vector on the user's machine. - EXTERNAL_DOWNLOADS (MEDIUM): The workflow involves searching GitHub and fetching repository contents (READMEs) to extract installation strings. It treats third-party, potentially adversarial content as authoritative instructions for system configuration.
- DATA_EXFILTRATION (MEDIUM): The skill requires permissions to read and write
~/.claude.json. This file often contains sensitive environment variables, authentication tokens, and API keys for other tools. An automated process modifying this file could inadvertently expose or corrupt these secrets.
Recommendations
- AI detected serious security threats
Audit Metadata