> ## Documentation Index
> Fetch the complete documentation index at: https://ocx.kdco.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Permissions

> Fine-grained access control for tools, bash commands, skills, and agent permissions

## Overview

Permissions control which actions require approval. Configure them globally, per-tool, or per-agent.

## Permission Values

| Value     | Behavior             |
| --------- | -------------------- |
| `"allow"` | Run without approval |
| `"ask"`   | Prompt for approval  |
| `"deny"`  | Disable the tool     |

## Default Permissions

Most operations are allowed by default. Exceptions:

* `doom_loop`: `ask`
* `external_directory`: `ask`

## Global Permissions

```jsonc theme={null}
{
  "permission": {
    "edit": "allow",
    "bash": "ask",
    "skill": "ask",
    "webfetch": "deny",
    "doom_loop": "ask",
    "external_directory": "ask"
  }
}
```

## Bash Command Permissions

Use glob patterns to control specific commands:

```jsonc theme={null}
{
  "permission": {
    "bash": {
      "git push": "ask",
      "git status": "allow",
      "terraform *": "deny",
      "*": "ask"
    }
  }
}
```

## Skill Permissions

```jsonc theme={null}
{
  "permission": {
    "skill": {
      "*": "deny",
      "git-*": "allow",
      "frontend/*": "ask"
    }
  }
}
```

## Per-Agent Permissions

Override global permissions for specific agents:

```jsonc theme={null}
{
  "permission": {
    "bash": { "git push": "ask" }
  },
  "agent": {
    "build": {
      "permission": {
        "bash": { "git push": "allow" }
      }
    }
  }
}
```

## Markdown Agent Permissions

```markdown theme={null}
---
description: Code review without edits
mode: subagent
permission:
  edit: deny
  bash:
    "git diff": allow
    "git log*": allow
    "*": ask
  webfetch: deny
---
```

## Best Practices

1. **Start restrictive, allow as needed** — Deny by default for sensitive operations.
2. **Use wildcards for grouped permissions** — `my-mcp*` controls all tools from an MCP server.
3. **Override per-agent when appropriate** — Give `build` agent more access than `plan`.
4. **Use `ask` for destructive operations** — `git push`, `rm`, deployment commands.

## See Also

* [Agents](/docs/reference/agents) — Agent configuration with permission overrides.
* [Skills & Instructions](/docs/reference/skills) — Skill access control.
* [Profiles Security](/docs/profiles/security) — Lock down recipes for untrusted repos.