self-improver

Warn

Audited by Socket on Apr 12, 2026

1 alert found:

Security
SecurityMEDIUM
SKILL.md

SUSPICIOUS: the core purpose is plausible, but the footprint is too powerful for a meta-skill. It can ingest untrusted skill instructions, run commands, mutate local skill files, install packages, and pull/reset from a third-party personal GitHub repo with author/source mismatch. Main concerns are supply-chain trust, autonomous modification, and indirect prompt injection rather than confirmed malware.

Confidence: 86%Severity: 78%
Audit Metadata
Analyzed At
Apr 12, 2026, 10:54 PM
Package URL
pkg:socket/skills-sh/mary4data%2FClawBee%2Fself-improver%2F@27d4697a623f383b0c4e4ce22d85774fb9a695ea