audit-permissions

Auditing permission grants involves scanning Claude Code configuration files to classify allow-list rules by destructive potential and credential exposure. This skill interactively helps you remove risky standing permissions to minimize security vulnerabilities.

Accumulated allow-list rules in agent settings create security risks like unauthorized system access and credential exposure. Forgotten or overly broad permission grants can allow destructive commands to execute without proper authorization.

Yes, Python 3 is required to execute the internal permission audit script. The skill relies on Python 3 to scan local Claude Code settings files and classify allow-list rules by risk level.

Yes, you can safely remove risky permissions without breaking your agent workflow. The skill uses an interactive human-in-the-loop cleanup process to narrow down or remove risky allow-list rules while maintaining your current agent flow.

Baseline verification detects permission drift by comparing your current allow-list rules against role-based baselines. This process identifies potential privilege escalation or security risks caused by accumulated or overly broad permission grants in your settings.