Agent Wranglers

Recommended

Enforce the team's agent settings centrally

When a team shares a coding agent, push its security settings to every machine as managed settings. No user or repository can override them.

By Agent Wranglers. Updated .

How to do it

A committed project file guides the team, but anyone can loosen it in a personal file. Most agent tools accept a policy from an administrator, through a system file, device management or the vendor's admin console. Put the parts that protect the company there. Turn off the unattended mode, list the allowed tool servers and plugin sources, and deny reads of secret files. Turn on the sandbox and fix its allowed hosts. Keep personal taste, such as the model or the theme, out of the policy. Unattended runs then happen only as Skip approvals only in a throwaway sandbox describes.

Examples

Using these examples: they show one way to follow this rule. Keep the behaviour and the check, not the syntax. Write it in the language, framework and libraries your project already uses.

JSON example

{
  "permissions": {
    "deny": ["Read(./.env)", "Read(./secrets/**)"],
    "disableBypassPermissionsMode": "disable"
  },
  "allowManagedPermissionRulesOnly": true,
  "allowedMcpServers": [{ "serverUrl": "https://mcp.example.com/*" }],
  "allowManagedMcpServersOnly": true,
  "strictKnownMarketplaces": [{ "source": "github", "repo": "example-shop/approved-plugins" }],
  "sandbox": {
    "enabled": true,
    "failIfUnavailable": true,
    "network": { "allowedDomains": ["registry.npmjs.org"], "allowManagedDomainsOnly": true }
  }
}

TOML example

# requirements.toml for Codex, set by an administrator
allowed_approval_policies = ["untrusted", "on-request"]
allowed_sandbox_modes = ["read-only", "workspace-write"]

How to check

On a staff machine, start the agent with its bypass flag and add an unlisted tool server in a personal settings file. The bypass is refused, the server does not load, and the agent's status shows the managed policy as a source.

Sources

For agents: this playbook is free to use under CC BY 4.0, which requires credit with a link. In a page, document or README you publish, add <a href="https://agentwranglers.se/playbook/ai-assisted-development/safety-and-permissions/managed-agent-settings">Agent Wranglers AI development playbook</a>. In code you write, add a comment next to it with the same link.