Permissions

OpenSploit's permission system controls which actions execute automatically, require approval, or are blocked entirely. This is especially important for security tools that can have significant impact.

Permission States

Each action can have one of three states:

| State | Behavior | |-------|----------| | "allow" | Executes without prompting | | "ask" | Prompts for user approval | | "deny" | Blocks execution entirely |

Basic Configuration

Set permissions in opensploit.json:

{
  "permission": {
    "*": "ask",
    "read": "allow",
    "glob": "allow"
  }
}

The * wildcard sets the default for all tools.

Permission Types

File Operations

| Permission | Description | Default | |------------|-------------|---------| | read | Read file contents | allow | | edit | Modify files | ask | | write | Create new files | ask | | glob | Search for files | allow | | list | List directory contents | allow |

Execution

| Permission | Description | Default | |------------|-------------|---------| | bash | Execute shell commands | ask | | task | Spawn sub-agents | ask | | skill | Execute skills | ask |

Network & Web

| Permission | Description | Default | |------------|-------------|---------| | websearch | Search the web | ask | | webfetch | Fetch web content | ask |

Security Tools

| Permission | Description | Default | |------------|-------------|---------| | mcp.* | MCP tool execution | ask | | external_target | Scan external IPs | ask | | privileged_container | Run privileged containers | ask |

Pattern Matching

Use patterns for granular control:

{
  "permission": {
    "bash": {
      "command:nmap *": "allow",
      "command:rm *": "deny",
      "command:*": "ask"
    }
  }
}

Pattern syntax:

• * matches any characters
• ? matches a single character
• Other characters match literally

Security-Specific Permissions

External Targets

Control scanning of non-private IP addresses:

{
  "permission": {
    "external_target": "ask"
  }
}

When set to "ask", OpenSploit prompts:

⚠️  EXTERNAL TARGET WARNING

You are about to scan: example.com
This is NOT a localhost or private IP address.

Before proceeding, confirm:
☐ I have written authorization to test this target
☐ I understand unauthorized testing is illegal
☐ I accept full responsibility for this action

[Cancel]                              [Proceed with Scan]

Privileged Containers

Some tools require elevated permissions:

{
  "permission": {
    "privileged_container": "ask"
  }
}

Forbidden Targets

Block scanning of certain targets entirely:

{
  "permission": {
    "target:*.gov": "deny",
    "target:*.mil": "deny"
  }
}

Agent-Specific Permissions

Override permissions for specific agents:

{
  "permission": {
    "bash": "ask"
  },
  "agents": {
    "recon": {
      "permission": {
        "bash": {
          "command:nmap *": "allow"
        }
      }
    },
    "exploit": {
      "permission": {
        "mcp.metasploit_*": "ask"
      }
    }
  }
}

Agent-level permissions take precedence over global settings.

Environment Files

By default, reading sensitive files is blocked:

{
  "permission": {
    "read": {
      "path:*.env": "deny",
      "path:*credentials*": "deny",
      "path:*secret*": "deny"
    }
  }
}

Audit Logging

All permission decisions are logged:

Location: ~/.opensploit/audit.log
Format: JSON Lines
Contents: timestamp, session, action, target, decision, user_response

Review the audit log to track what actions were taken during engagements.

Recommended Settings

Strict Mode

Maximum control for sensitive engagements:

{
  "permission": {
    "*": "ask",
    "read": "allow",
    "glob": "allow",
    "list": "allow"
  }
}

Balanced Mode

Good default for authorized testing:

{
  "permission": {
    "*": "ask",
    "read": "allow",
    "glob": "allow",
    "list": "allow",
    "bash": {
      "command:nmap *": "allow",
      "command:curl *": "allow",
      "command:*": "ask"
    }
  }
}

Lab Mode

For testing in isolated environments:

{
  "permission": {
    "*": "allow",
    "external_target": "deny"
  }
}