Share 
## https://sploitus.com/exploit?id=5519993A-3EC9-50C4-9972-B2C95A3394C4
# Red GSD v5.1 β€” Autonomous Red Team Agent Engine

> `red-gsd ` β€” one command. Full autonomous pipeline: recon β†’ reverse β†’ solve β†’ exploit β†’ verify β†’ report + scorecard.

## Architecture

```
red-gsd 
  └─ engine.py          # Main loop: plan β†’ dispatch β†’ collect β†’ decide β†’ loop
       β”œβ”€ scheduler.py   # Decision engine: phase transitions, LLR-based confidence
       β”œβ”€ runner.py      # Skill executor: internal stubs + external subprocess
       β”œβ”€ engage_schema.py # JSON state schema (ENGAGE.json)
       └─ report_builder.py # Auto-generated reports + 0-10 scorecard
```

## Quick Start

```bash
# Install
pip install -r engagement/requirements.txt

# Run
./bin/red-gsd 
```

## Skills (14 modules)

| Skill | Phase | Description |
|-------|-------|-------------|
| `new-engagement` | Init | Deep-dive target scoping, writes SCOPE.md |
| `strategy` | Plan | Attack strategy & tactical metrics |
| `roadmap` | Plan | Attack roadmap generation |
| `research` | Discovery | Target tech stack intel research |
| `plan` | Plan | Phased attack plan design |
| `recon` | Recon | nmap full-port + service fingerprint + web discovery |
| `reverse` | Reverse | Binary static/dynamic analysis β†’ vulnerability modeling |
| `exploit` | Exploit | CVE matching + exploit development |
| `verify` | Verify | Flag capture + evidence collection |
| `report` | Report | Auto-generated engagement report |
| `scorecard` | Report | 0-10 quantitative scoring |
| `transition` | Control | Phase auto-advancement & closed-loop control |
| `autonomous` | Meta | 12-skill fully autonomous orchestration |
| `build` | Meta | GSD design: Fable 5 design + Gemini 3.5 execution + Fable review |

## Configuration

Copy `config.example.json` to `~/.hermes/skills/red-teaming/red-gsd-config.json`:

| Key | Purpose |
|-----|---------|
| `pipeline.max_feedback_rounds` | Max retry attempts per phase |
| `pipeline.min_confidence` | Reverse→Solver confidence threshold (default 0.4) |
| `pipeline.exploit_timeout_minutes` | External skill subprocess timeout (seconds) |
| `llr.defaults` | LLR calibration table for observed_signals |
| `scoring.*` | Scorecard dimensions & thresholds |

## Testing

```bash
cd engagement
make test        # Unit + E2E tests
make v5-polish   # Full v5 test suite
```

## License

MIT