aa15d95986
Adds lightweight AI governance extension with: - R6 workflow formalism (Rules, Role, Request, Reference, Resource → Result) - Session start/end hooks for session tracking and audit initialization - Before/after tool call hooks for R6 request logging and audit trails - /audit command for generating governance reports - Hash-linked audit chain for verifiable provenance The extension provides transparent accountability for AI agent actions without impeding productivity, using a policy-based approach that can be customized per team or organization. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
180 lines
4.6 KiB
Markdown
180 lines
4.6 KiB
Markdown
# Web4 Governance Extension for Moltbot
|
|
|
|
Lightweight AI governance with R6 workflow formalism and audit trails.
|
|
|
|
[](https://opensource.org/licenses/MIT)
|
|
|
|
## Overview
|
|
|
|
This extension adds structured governance to Moltbot sessions:
|
|
|
|
- **R6 Workflow** - Every tool call follows a formal intent→action→result flow
|
|
- **Audit Trail** - Verifiable chain of actions with provenance
|
|
- **Session Identity** - Software-bound tokens for session tracking
|
|
|
|
No external dependencies. No network calls. Just structured, auditable AI actions.
|
|
|
|
## Installation
|
|
|
|
The extension is bundled with Moltbot. To enable it, configure in your Moltbot config:
|
|
|
|
```json
|
|
{
|
|
"plugins": {
|
|
"web4-governance": {
|
|
"enabled": true,
|
|
"auditLevel": "standard",
|
|
"showR6Status": true,
|
|
"actionBudget": null
|
|
}
|
|
}
|
|
}
|
|
```
|
|
|
|
## What It Does
|
|
|
|
### Every Tool Call Gets an R6 Record
|
|
|
|
The R6 framework captures structured intent:
|
|
|
|
```
|
|
R6 = Rules + Role + Request + Reference + Resource → Result
|
|
```
|
|
|
|
| Component | What It Captures |
|
|
|-----------|------------------|
|
|
| **Rules** | Preferences and constraints |
|
|
| **Role** | Session identity, action index |
|
|
| **Request** | Tool name, category, target |
|
|
| **Reference** | Chain position, previous R6 |
|
|
| **Resource** | (Optional) Estimated cost |
|
|
| **Result** | Status, output hash |
|
|
|
|
### Audit Trail with Provenance
|
|
|
|
Each action creates an audit record linked to its R6 request:
|
|
|
|
```json
|
|
{
|
|
"record_id": "audit:f8e9a1b2",
|
|
"r6_request_id": "r6:f8e9a1b2",
|
|
"tool": "Edit",
|
|
"category": "write",
|
|
"target": "src/main.rs",
|
|
"result": {
|
|
"status": "success",
|
|
"output_hash": "a1b2c3d4..."
|
|
},
|
|
"provenance": {
|
|
"session_id": "abc123",
|
|
"action_index": 47,
|
|
"prev_record_hash": "..."
|
|
}
|
|
}
|
|
```
|
|
|
|
Records form a hash-linked chain, enabling verification.
|
|
|
|
### Session Identity
|
|
|
|
Sessions get a software-bound token:
|
|
|
|
```
|
|
web4:session:a1b2c3d4
|
|
```
|
|
|
|
This is **not** hardware-bound (no TPM/Secure Enclave). Trust interpretation is up to the relying party. For hardware-bound identity, see [Hardbound](https://github.com/dp-web4/hardbound).
|
|
|
|
## Commands
|
|
|
|
| Command | Description |
|
|
|---------|-------------|
|
|
| `/audit` | Show session audit summary |
|
|
| `/audit last 10` | Show last 10 actions |
|
|
| `/audit verify` | Verify chain integrity |
|
|
| `/audit export` | Export audit log |
|
|
|
|
## Configuration
|
|
|
|
Available configuration options:
|
|
|
|
```json
|
|
{
|
|
"plugins": {
|
|
"web4-governance": {
|
|
"auditLevel": "standard",
|
|
"showR6Status": true,
|
|
"actionBudget": null
|
|
}
|
|
}
|
|
}
|
|
```
|
|
|
|
**auditLevel**:
|
|
- `minimal` - Just record, no output
|
|
- `standard` - Session start message
|
|
- `verbose` - Show each R6 request
|
|
|
|
**showR6Status**: Show session token on session start
|
|
|
|
**actionBudget**: Maximum number of actions (null = unlimited)
|
|
|
|
## Files
|
|
|
|
```
|
|
~/.web4/
|
|
├── preferences.json # User preferences (optional)
|
|
├── sessions/ # Session state
|
|
│ └── {session_id}.json
|
|
├── audit/ # Audit records
|
|
│ └── {session_id}.jsonl
|
|
└── r6/ # R6 request logs
|
|
└── {date}.jsonl
|
|
```
|
|
|
|
## Why R6?
|
|
|
|
The R6 framework provides:
|
|
|
|
1. **Structured Intent** - Every action has documented purpose
|
|
2. **Audit Foundation** - Machine-readable action history
|
|
3. **Context Preservation** - Reference links maintain history
|
|
4. **Trust Basis** - Verifiable record for trust evaluation
|
|
5. **Policy Hook** - Rules component enables future enforcement
|
|
|
|
R6 is observational by default - it records, doesn't block. This makes it safe to deploy without disrupting workflows.
|
|
|
|
## Web4 Ecosystem
|
|
|
|
This extension implements a subset of the [Web4 trust infrastructure](https://github.com/dp-web4/web4):
|
|
|
|
| Concept | This Extension | Full Web4 |
|
|
|---------|----------------|-----------|
|
|
| Identity | Software token | LCT (hardware-bound) |
|
|
| Workflow | R6 framework | R6 + Policy enforcement |
|
|
| Audit | Hash-linked chain | Distributed ledger |
|
|
| Trust | (Relying party decides) | T3 Trust Tensor |
|
|
|
|
For enterprise features (hardware binding, team governance, policy enforcement), see [Hardbound](https://github.com/dp-web4/hardbound).
|
|
|
|
## Contributing
|
|
|
|
Contributions welcome! This extension is MIT licensed.
|
|
|
|
Areas for contribution:
|
|
- Additional audit visualizations
|
|
- R6 analytics and insights
|
|
- Integration with external audit systems
|
|
- Performance optimizations
|
|
|
|
## License
|
|
|
|
MIT License - see [LICENSE](LICENSE)
|
|
|
|
## Links
|
|
|
|
- [Web4 Specification](https://github.com/dp-web4/web4)
|
|
- [R6 Framework Spec](https://github.com/dp-web4/web4/blob/main/web4-standard/core-spec/r6-framework.md)
|
|
- [Hardbound (Enterprise)](https://github.com/dp-web4/hardbound)
|
|
- [Moltbot](https://github.com/moltbot/moltbot)
|