review-permissions

To audit Claude Code permissions.allow rules for command injection risks, enumerate each glob entry in .claude/settings.json and assess it against a structured security checklist targeting flag abuse, path traversal, and sensitive file reads. The review generates per-rule findings with concrete fixes.

Overly-permissive Claude Code allowlist rules can cause command injection, secret exposure, and privilege escalation. Evaluating allow rules against a targeted attack checklist identifies these vulnerabilities before they enable unauthorized access to sensitive files or system commands.

Detecting dangerous cross-rule combinations in Claude Code security settings requires testing multi-rule chains rather than assessing rules in isolation. This approach flags sequences like read-then-exfiltrate that appear safe individually but enable data theft when combined.

You can safely grant privileged script access in Claude Code by following a three-layer strict-shape pattern: exact allow entries in settings, a dedicated PreToolUse:Bash guard, and script-internal argument validation. This prevents prompt-manipulated command crafting.

You should review your .claude/settings.json permissions for secret protection whenever permissions.allow entries are added or modified. This ensures new glob rules do not inadvertently enable path traversal to sensitive files or permit unauthorized secret access.

The Claude Code permissions review provides concrete fixes for insecure allow rules. It reports per-rule findings along with dangerous cross-rule combinations, offering tighter replacement glob entries to prevent injection and privilege escalation.