False positive: substreams-search-mcp flagged as suspicious

[PaulieB14]

Skill

• Slug: substreams-search-mcp
• Owner: @PaulieB14
• URL: https://clawhub.ai/PaulieB14/substreams-search-mcp
• Version: 1.3.0 (1.3.1 pending publish due to rate limit)
• Source: https://github.com/PaulieB14/substreams-search-mcp

What the skill does

This is an MCP server that lets AI agents search, inspect, and analyze Substreams packages from the substreams.dev registry. It exposes 4 tools: search_substreams, inspect_package, list_package_modules, and get_sink_config.

Expected flags and why they are false positives

1. "Environment variable access combined with network send" — The server reads process.env.MCP_HTTP_PORT to configure the optional SSE/HTTP transport port. This is the standard pattern for any MCP server supporting dual transport (stdio + SSE) per the MCP SDK.

2. npx launch mechanism — Standard for npm-published MCP servers, recommended by Claude Desktop, Cursor, Claude Code, and OpenClaw.

3. Network requests to substreams.dev and spkg.io — The search_substreams tool scrapes the public substreams.dev registry. The inspect_package and get_sink_config tools fetch .spkg files from spkg.io to parse protobuf metadata. All public APIs, no API key required. Documented in SKILL.md.

4. HTTP/SSE transport on local port — Opt-in via --http or --http-only CLI flags for OpenClaw/remote agent compatibility. Default port 3849, configurable via MCP_HTTP_PORT.

Why this is a false positive

• VirusTotal should mark this as Benign
• No API keys, secrets, or credentials are required or collected
• All network requests go to public registries (substreams.dev, spkg.io)
• No local database or persistent storage
• SSE transport is opt-in and uses the standard MCP SDK SSEServerTransport
• All behavior is documented in SKILL.md under "Requirements" and "Network & Data Behavior"