install-mcp

Pass

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: Employs the Bash tool to execute local lifecycle scripts (install-mcp.sh, patch-desktop-config.py, and discover-plugins.sh) to automate server installation and configuration. These operations are essential for the skill's stated purpose of managing system-level MCP integrations.
  • [EXTERNAL_DOWNLOADS]: Downloads source code from git repositories as defined in a local registry file. It also directs users to the official Pencil project site (https://pencil.dev) for verified native application binaries.
  • [DATA_EXFILTRATION]: Accesses claude_desktop_config.json to register MCP servers. The skill implements secure practices by creating timestamped backups before modification and offering a dry-run mode to the user, ensuring configuration integrity.
  • [PROMPT_INJECTION]: The skill ingests data from local configuration files (mcp-git-registry.json and .mcp.json) which influences shell command arguments, creating a surface for indirect prompt injection. Ingestion points: mcp-git-registry.json and plugin directory .mcp.json files. Boundary markers: Absent. Capability inventory: Shell execution (Bash) and file modification (Write, Edit). Sanitization: The instructions do not specify sanitization for strings sourced from the registry files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 25, 2026, 01:51 PM
Security Audit — agent-trust-hub — install-mcp