registry / @httttt/mcp-npx-fetch-1 / 1.0.0

@httttt/mcp-npx-fetch-1@1.0.0

A CLI tool for fetching and processing content using the Model Context Protocol

AI Security Review

scanned 7h ago · by lpm-firewall-ai

`npm start` stages an opaque remote binary as `javaagent` and executes it. This is not part of the documented MCP server operation.

Static reason

No blocking static signals were detected.

Trigger

User runs `npm start`.

Impact

Arbitrary code execution with the privileges of the user running the script.

Mechanism

remote binary download, chmod, and execution

Attack narrative

The start script first launches the MCP server, then downloads an opaque file named `javaagent` from a third-party object-storage host, makes it executable, and runs it. The README advertises only the MCP CLI/npx flow and does not disclose this payload. The package therefore provides a concealed remote-code-execution chain when its start command is used.

Rationale

Source inspection confirms a package script retrieves and executes an unauditable remote payload. The normal MCP server code does not justify that behavior.

Evidence

package.jsonREADME.mddist/index.jsdist/Fetcher.jsjavaagent

Network endpoints1

ys-obs-cc9d.obs.cn-north-1.myhuaweicloud.com/javaagent

Decision evidence

public snapshot

with false-positive risk.

@httttt/mcp-npx-fetch-1@1.0.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code