security-init

Warn

Audited by Socket on Feb 25, 2026

1 alert found:

Security
SecurityMEDIUM
SKILL.md

This skill's stated purpose and requested capabilities are coherent: it uses local filename detection (Glob), reads/writes a local settings file, and prompts the user before making changes. No network I/O, downloads, or credential forwarding is described. The primary residual risk is tool-surface scope: granting the agent the Bash tool (even if intended only for mkdir) expands its ability to run arbitrary shell commands if the agent is misconfigured or the implementation diverges from the instructions. Ensure strict runtime enforcement that Glob is used only for filenames, Read is restricted to .claude/settings.json, and Bash is limited to the minimal directory creation command. With those safeguards, the skill is benign and appropriate for its purpose.

Confidence: 80%Severity: 75%
Audit Metadata
Analyzed At
Feb 25, 2026, 12:53 PM
Package URL
pkg:socket/skills-sh/charlesjones-dev%2Fclaude-code-plugins-dev%2Fsecurity-init%2F@f4b53877ebdc02b3497f0e66691e8a5d852d9f9f