Agent IndexAgent Permissions Index

Grade my agent config

Paste your .claude/settings.json, .gemini/settings.json or mcp.json. We count grades anonymously (tool + letter only); your config never leaves your browser.

Tip: secrets are only checked locally. Still, consider redacting before pasting.

What we check
The same signals measured across thousands of public configs.
Bypass / YOLO mode
Unrestricted shell
Wildcard interpreters (python, node, npx…)
rm and curl wildcards
Missing deny rules
Agent can edit its own permissions
Hardcoded credential-looking values