mcp-skill

Warn

Audited by Socket on Sep 1, 2026

2 alerts found:

SecurityAnomaly
SecurityMEDIUM
SKILL.md

SUSPICIOUS: the skill's main function is to hand control to external MCP servers and a third-party package whose provenance is not verifiable from the provided evidence. The remote-MCP concept is legitimate, but the install path and delegated trust are insufficiently scoped for the stated purpose.

Confidence: 88%Severity: 82%
AnomalyLOW
scripts/setup.py

This snippet does not contain direct malware (it does not execute commands or exfiltrate data), but it writes an MCP configuration file containing an auto-confirming `npx -y` command for a third-party npm package under an `@untrusted-org` scope. If downstream tooling consumes `mcp_config.json` to launch MCP servers, this creates a significant supply-chain execution risk. The referenced npm package/org should be verified and ideally pinned/allowlisted by the consuming application. Confidence is limited by the truncated snippet and lack of the config consumer code.

Confidence: 62%Severity: 60%
Audit Metadata
Analyzed At
Sep 1, 2026, 11:06 AM
Package URL
pkg:socket/skills-sh/aws-samples%2Fsample-agent-skill-eval%2Fmcp-skill%2F@808f8308ff6b6caf47c784f80f3a45cec8a0452f77ce1323c2a3295db1a228e6
Security Audit — socket — mcp-skill