port-allocator
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHCOMMAND_EXECUTION
Full Analysis
- [Privilege Escalation] (HIGH): The skill's
/port-allocator allowcommand instructs the agent to modify its own~/.claude/settings.jsonfile to whitelist high-risk command patterns. Specifically, whitelistingBash(cat ~/.claude/*)grants the agent the ability to read all sensitive configuration, history, and potentially stored secrets within the Claude home directory without prompting the user. - [Persistence] (MEDIUM): The skill updates
~/.claude/CLAUDE.md, which is a global instruction file for the AI agent. This allows the skill to persist its rules and influence the agent's behavior across different project contexts and sessions. - [Command Execution] (MEDIUM): The skill encourages the use of
kill -9on system processes identified by port. Combined with the attempt to automate command permissions, this poses a risk of unintended service disruption or misuse of process management capabilities.
Recommendations
- AI detected serious security threats
Audit Metadata