yovinchen/claudia
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor: remove sandbox system and simplify agent architecture

Browse Source 
Remove the entire sandbox security system including:
- All sandbox-related Rust code and dependencies (gaol crate)
- Sandbox command handlers and platform-specific implementations
- Comprehensive test suite for sandbox functionality
- Agent sandbox settings UI components

Simplify agent configuration by removing sandbox and permission fields:
- Remove sandbox_enabled, enable_file_read, enable_file_write, enable_network from agent configs
- Update all CC agents to use simplified configuration format
- Remove sandbox references from documentation and UI
This commit is contained in:
Vivek R
2025-07-02 19:17:38 +05:30
parent 124fe1544f
commit 2dfdf31b83
47 changed files with 115 additions and 7774 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
src-tauri/tests/TESTS_COMPLETE.md
View File

@@ -21,7 +21,7 @@ test result: ok. 58 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
### Implementation Details:
#### Real Claude Execution (`tests/sandbox/common/claude_real.rs`):
#### Real Claude Execution:
- `execute_claude_task()` - Executes Claude with specified task and captures output
- Supports timeout handling (gtimeout on macOS, timeout on Linux)
- Returns structured output with stdout, stderr, exit code, and duration
@@ -33,18 +33,18 @@ test result: ok. 58 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
- 20-second timeout to allow Claude sufficient time to respond
#### Key Test Updates:
1. **Agent Tests** (`agent_sandbox.rs`):
- `test_agent_with_minimal_profile` - Tests with minimal sandbox permissions
- `test_agent_with_standard_profile` - Tests with standard permissions
- `test_agent_without_sandbox` - Control test without sandbox
1. **Agent Tests**:
- Test agent execution with various permission configurations
- Test agent execution in different project contexts
- Control tests for baseline behavior
2. **Claude Sandbox Tests** (`claude_sandbox.rs`):
- `test_claude_with_default_sandbox` - Tests default sandbox profile
- `test_claude_sandbox_disabled` - Tests with inactive sandbox
2. **Claude Tests**:
- Test Claude execution with default settings
- Test Claude execution with custom configurations
### Benefits of Real Claude Testing:
- **Authenticity**: Tests validate actual Claude behavior, not mocked responses
- **Integration**: Ensures the sandbox system works with real Claude execution
- **Integration**: Ensures the system works with real Claude execution
- **End-to-End**: Complete validation from command invocation to output parsing
- **No External Dependencies**: Uses `--dangerously-skip-permissions` flag
@@ -53,6 +53,6 @@ test result: ok. 58 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
- No ignored tests
- No TODOs in test code
- Clean compilation with no warnings
- Platform-aware sandbox expectations (Linux vs macOS)
- Platform-aware expectations for different operating systems
The test suite now provides comprehensive end-to-end validation with actual Claude execution.
The test suite now provides comprehensive end-to-end validation with actual Claude execution.