understanding-tauri-runtime-authority
npx skills add https://github.com/beshkenadze/claude-code-tauri-skills --skill understanding-tauri-runtime-authority
Agent 安装分布
Skill 文档
Tauri Runtime Authority
The runtime authority is a core Tauri component that enforces security policies during application execution. It validates permissions, resolves capabilities, and injects scopes before commands execute.
What Is Runtime Authority?
Runtime authority is the enforcement layer that sits between the WebView frontend and Tauri commands. It acts as a gatekeeper for all IPC (Inter-Process Communication) requests.
Core Function
When a webview invokes a Tauri command, the runtime authority:
- Receives the invoke request from the webview
- Validates the origin is permitted to call the requested command
- Confirms the origin belongs to applicable capabilities
- Injects defined scopes into the request
- Passes the validated request to the Tauri command
If the origin is not allowed, the request is denied and the command never executes.
Trust Boundary Model
Tauri implements a trust boundary separating Rust core code from WebView frontend code:
| Zone | Trust Level | Access |
|---|---|---|
| Rust Core | Full trust | Unrestricted system access |
| WebView Frontend | Limited trust | Only exposed resources via IPC |
The runtime authority enforces this boundary at execution time.
Security Architecture
How Runtime Authority Fits
Frontend (WebView)
|
v
[IPC Invoke Request]
|
v
+------------------+
| Runtime Authority| <-- Validates permissions, capabilities, scopes
+------------------+
|
v (if allowed)
[Tauri Command Execution]
|
v
[System Resources]
Key Components
| Component | Role in Runtime |
|---|---|
| Permissions | Define what commands exist and their access rules |
| Capabilities | Map permissions to specific windows/webviews |
| Scopes | Restrict command behavior with path/resource limits |
| Runtime Authority | Enforces all of the above at execution time |
Capability Resolution at Runtime
When a command is invoked, the runtime authority resolves which capabilities apply.
Resolution Process
- Identify Origin: Determine which window/webview made the request
- Match Capabilities: Find all capabilities that include this window
- Collect Permissions: Aggregate all permissions from matched capabilities
- Check Command Access: Verify the command is allowed
- Merge Scopes: Combine all applicable scope restrictions
- Validate or Deny: Either proceed with scope injection or reject
Window Capability Merging
When a window is part of multiple capabilities, security boundaries merge:
// capability-1.json
{
"identifier": "basic-access",
"windows": ["main"],
"permissions": ["fs:allow-read-file"]
}
// capability-2.json
{
"identifier": "write-access",
"windows": ["main"],
"permissions": ["fs:allow-write-file"]
}
Result: The “main” window gets both read and write permissions.
Platform-Specific Resolution
Capabilities can target specific platforms. At runtime, only capabilities matching the current platform are considered:
{
"identifier": "desktop-features",
"platforms": ["linux", "macOS", "windows"],
"windows": ["main"],
"permissions": ["shell:allow-execute"]
}
On iOS/Android, this capability is ignored at runtime.
Access Control Enforcement
Deny Precedence Rule
When evaluating access, deny rules always take precedence:
{
"permissions": [
{
"identifier": "fs:allow-read-file",
"allow": [{ "path": "$HOME/**" }],
"deny": [{ "path": "$HOME/.ssh/**" }]
}
]
}
At runtime:
- Request to read
$HOME/documents/file.txt– Allowed - Request to read
$HOME/.ssh/id_rsa– Denied (deny rule matches)
Command-Level Validation
Before any command executes:
- Runtime authority checks if the command permission exists
- Verifies the calling window has that permission via its capabilities
- Validates any scope restrictions are satisfied
Window "editor" calls fs.readFile("/home/user/doc.txt")
|
v
Runtime Authority checks:
- Does "editor" have fs:allow-read-file? Yes
- Is "/home/user/doc.txt" in allowed scope? Yes
- Is it in any deny scope? No
|
v
Command executes with scopes injected
Scope Injection
How Scopes Work at Runtime
Scopes are not just validation rules; they are injected into command execution context. Commands can access their applicable scopes to enforce restrictions.
Scope Variables
At runtime, scope variables resolve to actual paths:
| Variable | Runtime Resolution |
|---|---|
$APP |
Application install directory |
$APPDATA |
App data directory |
$APPCONFIG |
App config directory |
$HOME |
User home directory |
$TEMP |
Temporary directory |
$DOCUMENT |
Documents directory |
$DOWNLOAD |
Downloads directory |
$DESKTOP |
Desktop directory |
Scope Combination Example
{
"identifier": "main-capability",
"windows": ["main"],
"permissions": [
{
"identifier": "fs:allow-read-file",
"allow": [{ "path": "$APPDATA/*" }]
},
{
"identifier": "fs:allow-write-file",
"allow": [{ "path": "$APPDATA/config.json" }]
}
]
}
At runtime for the “main” window:
- Read operations allowed in
$APPDATA/* - Write operations only allowed for
$APPDATA/config.json
Path Traversal Prevention
The runtime authority includes built-in path traversal protection:
Request: /usr/path/to/../../../etc/passwd
Result: DENIED (path traversal detected)
Parent directory accessors (..) in paths are blocked, ensuring scope restrictions cannot be bypassed.
Configuration Examples
Basic Runtime Security Setup
src-tauri/capabilities/default.json:
{
"$schema": "../gen/schemas/desktop-schema.json",
"identifier": "default-capability",
"description": "Default runtime permissions",
"windows": ["main"],
"permissions": [
"core:default",
"core:event:default",
"core:window:default"
]
}
Scoped Filesystem Access
src-tauri/capabilities/files.json:
{
"$schema": "../gen/schemas/desktop-schema.json",
"identifier": "file-access",
"description": "Controlled filesystem access",
"windows": ["main"],
"permissions": [
"fs:default",
{
"identifier": "fs:allow-read-file",
"allow": [
{ "path": "$APPDATA/**" },
{ "path": "$DOCUMENT/**" }
],
"deny": [
{ "path": "$DOCUMENT/private/**" }
]
},
{
"identifier": "fs:allow-write-file",
"allow": [
{ "path": "$APPDATA/**" }
]
}
]
}
Multi-Window Security Boundaries
src-tauri/capabilities/editor.json:
{
"$schema": "../gen/schemas/desktop-schema.json",
"identifier": "editor-capability",
"description": "Full editor permissions",
"windows": ["editor"],
"permissions": [
"core:default",
"fs:default",
"fs:allow-read-file",
"fs:allow-write-file",
"dialog:default"
]
}
src-tauri/capabilities/preview.json:
{
"$schema": "../gen/schemas/desktop-schema.json",
"identifier": "preview-capability",
"description": "Read-only preview permissions",
"windows": ["preview"],
"permissions": [
"core:window:default",
"core:event:default",
{
"identifier": "fs:allow-read-file",
"allow": [{ "path": "$TEMP/preview/**" }]
}
]
}
At runtime:
- “editor” window can read/write files and open dialogs
- “preview” window can only read from temp preview directory
HTTP Request Scoping
{
"identifier": "api-access",
"windows": ["main"],
"permissions": [
{
"identifier": "http:default",
"allow": [
{ "url": "https://api.myapp.com/*" },
{ "url": "https://cdn.myapp.com/*" }
],
"deny": [
{ "url": "https://api.myapp.com/admin/*" }
]
}
]
}
Runtime Security Guarantees
What Runtime Authority Protects
| Threat | Protection |
|---|---|
| Frontend compromise | Limits damage to granted permissions only |
| Unauthorized command access | Commands denied without explicit capability |
| Path traversal attacks | Built-in prevention at runtime |
| Scope bypass attempts | All scopes enforced before command execution |
| Cross-window access | Each window isolated to its capabilities |
What Runtime Authority Does NOT Protect
| Threat | Why Not Protected |
|---|---|
| Malicious Rust code | Rust core has full trust |
| Overly permissive config | Developer responsibility |
| WebView vulnerabilities | OS WebView security boundary |
| Supply chain attacks | Dependency security |
Debugging Runtime Authority
Permission Denied Errors
When a command fails with permission denied:
- Check Window Label: Verify the window making the request
- Check Capability: Ensure a capability targets that window
- Check Permission: Verify the permission is in the capability
- Check Scope: Verify the resource is in allowed scope and not denied
Common Runtime Issues
| Issue | Cause | Solution |
|---|---|---|
| Command not allowed | Missing permission in capability | Add permission to capability |
| Scope not applied | No scope defined for permission | Add allow scope |
| Access denied despite allow | Deny rule takes precedence | Remove conflicting deny |
| Window has no permissions | Capability not targeting window | Check window label in capability |
Verifying Runtime Configuration
Check generated schemas to see what permissions are available:
src-tauri/gen/schemas/desktop-schema.json
src-tauri/gen/schemas/mobile-schema.json
Best Practices
Principle of Least Privilege
Grant only the permissions each window actually needs:
// Good: Specific permissions
{
"windows": ["settings"],
"permissions": [
"core:window:allow-close",
"fs:allow-read-file"
]
}
// Avoid: Overly broad permissions
{
"windows": ["settings"],
"permissions": ["fs:default", "shell:default"]
}
Always Define Scopes
Never leave filesystem or network permissions unscoped:
// Good: Scoped access
{
"identifier": "fs:allow-read-file",
"allow": [{ "path": "$APPDATA/**" }]
}
// Avoid: Unscoped access
{
"permissions": ["fs:allow-read-file"]
}
Deny Sensitive Paths
Explicitly deny access to sensitive locations:
{
"identifier": "fs:allow-read-file",
"allow": [{ "path": "$HOME/**" }],
"deny": [
{ "path": "$HOME/.ssh/**" },
{ "path": "$HOME/.gnupg/**" },
{ "path": "$HOME/.aws/**" }
]
}
Separate Capabilities by Trust Level
Create distinct capabilities for different security contexts:
capabilities/
main-trusted.json # Full access for main window
plugin-limited.json # Restricted for plugin windows
preview-readonly.json # Read-only for preview
Platform-Specific Security
Use platform targeting for OS-specific permissions:
{
"identifier": "desktop-shell",
"platforms": ["linux", "macOS", "windows"],
"windows": ["main"],
"permissions": ["shell:allow-execute"]
}
This prevents desktop-only permissions from being evaluated on mobile.
Summary
The runtime authority is Tauri’s enforcement mechanism for the ACL-based security model:
- Every IPC request passes through runtime authority validation
- Capabilities resolve at runtime based on the calling window
- Scopes inject into command execution context
- Deny rules always take precedence over allow rules
- Path traversal is blocked automatically
- Security boundaries merge when windows have multiple capabilities
Configure capabilities and permissions correctly, and the runtime authority ensures they are enforced consistently throughout application execution.