Permissions and Safety

Why Permissions Matter

AI agents can execute code, modify files, and run shell commands. Without guardrails, a prompt injection or mistake could delete files, leak secrets, or run malicious code. Permissions ensure you stay in control.

Claude Code's Built-in Permissions

Claude Code uses settings files to define what the agent can do:

• .claude/settings.json - Project settings (shared with team)

• .claude/settings.local.json - Personal project settings (gitignored)

• ~/.claude/settings.json - Global user settings

Each file can specify:

• allow - Patterns that auto-approve

• deny - Patterns that always block

• additionalDirectories - Folders outside the project the agent can access

Nimbalyst supports this.

What Nimbalyst Adds

In addition, Nimbalyst adds a workspace trust layer on top of the coding agent's permissions:

1. Trust Gate - Projects must be explicitly trusted before the agent can do anything

2. Permission Modes - Choose how much autonomy to grant the agent

3. Inline Confirmations - Approve or deny actions as they happen

Permission Modes

Mode	Behavior
Ask (Recommended)	Agent requests permission for each action. You can approve once, for the session, or permanently.
Allow All	File operations (Read, Write, Edit) auto-approve. Bash commands still require approval based on your settings.
Bypass All	Everything auto-approves. Use only in trusted, sandboxed environments.

Allowing All Web Searches while in Allow All Edits mode

To allow your coding agent to search and fetch all websites, in the advanced settings, add . on the URL allow list. You can also use wildcard domain patterns like *.github.com to allow all subdomains of a specific site, or use the "Allow All Domains" button for quick access to permit all web fetches.

How Approval Works in Ask Mode

When the agent wants to perform an action in Ask mode:

1. An inline confirmation appears with the action details

2. You choose:

• Deny - Block this request

• Allow Once - Allow just this time

• Session - Allow this pattern for the rest of the session

• Always - Save the pattern to .claude/settings.local.json

Managing Permissions

Open Settings > Agent Permissions to:

• Change your permission mode

• View and remove approved patterns

• Add additional directories

• Reset to defaults

Pattern Examples

• Bash(git:*) - Allow git commands

• Bash(npm:*) - Allow npm commands

• Edit - Allow file edits

• WebFetch(domain:github.com) - Allow fetching from github.com

Security Notes

• Sensitive paths (~/.ssh, ~/.aws, etc.) are always blocked

• Compound bash commands get one-time patterns that don't persist

• Untrusted projects deny all agent actions until you set a permission mode