agent-reach

Warn

Audited by Socket on Sep 28, 2026

1 alert found:

Security
SecurityMEDIUM
SKILL.md

SUSPICIOUS. The skill’s broad web-reading purpose matches its capabilities, but its install and trust model are weak: it installs from an unpinned GitHub branch archive, then expands into multiple external tools and credentialed integrations. Cookie extraction, proxy configuration, and third-party Exa MCP routing are proportionate to the stated purpose but increase exposure. This looks more like a risky, over-broad internet aggregation tool than clear malware.

Confidence: 90%Severity: 74%
Audit Metadata
Analyzed At
Sep 28, 2026, 02:29 PM
Package URL
pkg:socket/skills-sh/gpttang%2Fskill-atlas%2Fagent-reach%2F@a3deb7c4c9a32a9af6d039fad0a4160a30946e0125106b636ffac845fae3cc31