@aurora-foundation/obsidian-next 0.2.0 → 0.3.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +86 -12
- package/dist/index.js +1497 -419
- package/package.json +2 -2
package/README.md
CHANGED
|
@@ -4,8 +4,8 @@
|
|
|
4
4
|
|
|
5
5
|
|
|
6
6
|
|
|
7
|
-
|
|
8
|
+
|
|
9
9
|
|
|
10
10
|
**Obsidian Next** is a professional, structured, and secure AI agent interface for the terminal. Built by **Aurora Labs** (a division of the **Aurora Foundation**) with a "Structure-First" architecture for rigorous, interactive, and safe user experiences.
|
|
11
11
|
|
|
@@ -30,19 +30,47 @@ The `workspace/` directory is a dedicated environment where **Polyoxy** is curre
|
|
|
30
30
|
- **Benchmarks**: Comprehensive safety and performance benchmarks are running. Results will be published soon.
|
|
31
31
|
- **Evaluation Goal**: The current workspace is used to stress-test the Auditor's ability to catch malicious patterns in a controlled environment.
|
|
32
32
|
|
|
33
|
-
## Security
|
|
33
|
+
## Security Features (v0.3.0)
|
|
34
34
|
|
|
35
|
-
|
|
35
|
+
Obsidian Next implements **Zero Trust AI Automation** with the following security layers:
|
|
36
36
|
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
37
|
+
### Implemented (v0.3.0-security)
|
|
38
|
+
|
|
39
|
+
1. **Rotating Key System** [NEW]
|
|
40
|
+
- Secure API key storage via macOS Keychain, Linux secret-tool, or encrypted file fallback
|
|
41
|
+
- Machine-specific key derivation (AES-256-GCM)
|
|
42
|
+
- Auto-rotation detection for long sessions
|
|
43
|
+
- Never stores plaintext keys in config files
|
|
44
|
+
|
|
45
|
+
2. **PII Redaction Engine** [NEW]
|
|
46
|
+
- Real-time redaction of sensitive data before sending to LLM
|
|
47
|
+
- 14 built-in patterns: email, phone, SSN, credit cards, AWS keys, API tokens, passwords, private keys, JWT
|
|
48
|
+
- Configurable per-pattern enable/disable
|
|
49
|
+
- Allowlist support for specific values
|
|
50
|
+
|
|
51
|
+
3. **Audit Logging** [NEW]
|
|
52
|
+
- Complete audit trail of all command executions
|
|
53
|
+
- File operation logging (read/write/edit/delete)
|
|
54
|
+
- Approval decision tracking
|
|
55
|
+
- JSON format for easy parsing, auto-rotation at 10MB
|
|
56
|
+
|
|
57
|
+
4. **Approval Enforcement** [FIXED]
|
|
58
|
+
- Commands requiring approval now properly block execution
|
|
59
|
+
- Safe mode enforces approval for all write operations
|
|
60
|
+
- No bypass possible through mode switching
|
|
61
|
+
|
|
62
|
+
5. **Sandbox Runtime**
|
|
63
|
+
- OS-level isolation via `@anthropic-ai/sandbox-runtime`
|
|
64
|
+
- Native fallbacks to `sandbox-exec` (macOS) and `firejail` (Linux)
|
|
65
|
+
|
|
66
|
+
### Roadmap
|
|
67
|
+
|
|
68
|
+
1. **Hardware-Level Sandboxing**:
|
|
41
69
|
- Integration with native OS hypervisors (Apple Virtualization Framework) for true VM isolation.
|
|
42
|
-
|
|
43
|
-
- Pre-flight auditing to strip Personally Identifiable Information (PII) before it hits the LLM context.
|
|
44
|
-
4. **Signed Execution**:
|
|
70
|
+
2. **Signed Execution**:
|
|
45
71
|
- Only allowing cryptographically signed tool definitions to run.
|
|
72
|
+
3. **Network Isolation**:
|
|
73
|
+
- Per-session network namespaces for complete network control.
|
|
46
74
|
|
|
47
75
|
## Documentation Directory
|
|
48
76
|
|
|
@@ -106,13 +134,59 @@ Obsidian Next can be run as a Model Context Protocol (MCP) server.
|
|
|
106
134
|
### Usage
|
|
107
135
|
|
|
108
136
|
```bash
|
|
109
|
-
# Set your API Key
|
|
137
|
+
# Set your API Key (or use /init to store securely)
|
|
110
138
|
export ANTHROPIC_API_KEY="sk-ant-..."
|
|
111
139
|
|
|
112
140
|
# Start the Agent
|
|
113
141
|
npm start
|
|
114
142
|
```
|
|
115
143
|
|
|
144
|
+
### Commands
|
|
145
|
+
|
|
146
|
+
| Command | Description |
|
|
147
|
+
|---------|-------------|
|
|
148
|
+
| `/settings` | Interactive settings menu (arrow keys + Enter) |
|
|
149
|
+
| `/mode` | Set execution mode (auto/plan/safe) |
|
|
150
|
+
| `/models` | Select AI model |
|
|
151
|
+
| `/status` | Show system status |
|
|
152
|
+
| `/cost` | Show session cost |
|
|
153
|
+
| `/undo` | Undo file changes |
|
|
154
|
+
| `/sandbox` | Toggle sandbox mode |
|
|
155
|
+
| `/clear` | Clear conversation |
|
|
156
|
+
| `/doctor` | Run diagnostics |
|
|
157
|
+
| `/exit` | Exit the CLI |
|
|
158
|
+
|
|
159
|
+
### Settings Menu
|
|
160
|
+
|
|
161
|
+
Access the interactive settings menu with `/settings`:
|
|
162
|
+
|
|
163
|
+
```
|
|
164
|
+
[*] Settings
|
|
165
|
+
> [1] Execution Mode Current: safe
|
|
166
|
+
[2] Security PII redaction, audit logging
|
|
167
|
+
[3] UI Preferences Syntax highlighting, colors
|
|
168
|
+
[4] Permissions Allow/deny lists
|
|
169
|
+
[5] Close Settings
|
|
170
|
+
|
|
171
|
+
Arrows: navigate | Enter: select/toggle | Esc: back
|
|
172
|
+
```
|
|
173
|
+
|
|
174
|
+
### Security Settings
|
|
175
|
+
|
|
176
|
+
| Setting | Default | Description |
|
|
177
|
+
|---------|---------|-------------|
|
|
178
|
+
| `security.piiRedaction` | `true` | Redact PII before sending to LLM |
|
|
179
|
+
| `security.auditLogging` | `true` | Log all commands to audit.log |
|
|
180
|
+
| `security.keyBackend` | `auto` | Key storage: auto/keychain/secret-tool/encrypted-file |
|
|
181
|
+
|
|
182
|
+
### Execution Modes
|
|
183
|
+
|
|
184
|
+
| Mode | Description |
|
|
185
|
+
|------|-------------|
|
|
186
|
+
| `safe` | (Default) Require approval for all write operations |
|
|
187
|
+
| `plan` | Read-only planning, approve plan before execution |
|
|
188
|
+
| `auto` | Execute all commands without confirmation |
|
|
189
|
+
|
|
116
190
|
## References & Standards
|
|
117
191
|
|
|
118
192
|
This project adheres to strict industry standards:
|