https://github.com/plwp/duplicat-rex
Agentic SaaS reverse-engineering and duplication engine
https://github.com/plwp/duplicat-rex
ai browser-automation developer-tools playwright reverse-engineering specification
Last synced: 2 months ago
JSON representation
Agentic SaaS reverse-engineering and duplication engine
- Host: GitHub
- URL: https://github.com/plwp/duplicat-rex
- Owner: plwp
- Created: 2026-04-06T08:18:36.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-04-06T14:34:54.000Z (3 months ago)
- Last Synced: 2026-04-06T15:37:37.480Z (3 months ago)
- Topics: ai, browser-automation, developer-tools, playwright, reverse-engineering, specification
- Language: Python
- Size: 238 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 7
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Duplicat-Rex
Agentic SaaS reverse-engineering engine. Gathers intelligence from every available source, synthesises structured specifications, and drives chief-wiggum's build pipeline to produce full-feature-parity clones.
## What This Repo Is
Duplicat-rex owns the **intelligence pipeline**: reconnaissance, spec synthesis, behavioral comparison, and gap analysis. It does not generate code directly — it produces structured specifications and GitHub issues that chief-wiggum's build pipeline (`/plan-epic`, `/architect`, `/implement-wave`, `/close-epic`) consumes.
Three repos in play:
1. **duplicat-rex** (this repo) — intelligence/recon engine
2. **chief-wiggum** — build engine (configured as skill source)
3. **output repo** (e.g. `plwp/abuello`) — the generated app, one per target
```mermaid
graph LR
DR[duplicat-rex
Intelligence Layer] -->|specs & issues| CW[chief-wiggum
Build Engine]
CW -->|implementation| OUT[abuello
Output App]
DR -->|compare| OUT
OUT -->|gaps| DR
```
## Tech Stack
- **Language**: Python 3.11+
- **CLI**: Typer
- **AI**: Multi-model (Claude, Codex, Gemini) with structured adjudication
- **Browser automation**: Browser-use + Playwright
- **Transcription**: whisper + ffmpeg + yt-dlp
- **Secrets**: System keychain (never env vars)
## Setup
```bash
# Clone and install
git clone https://github.com/plwp/duplicat-rex
cd duplicat-rex
pip install -e ".[dev]"
# Configure secrets (API keys stored in macOS Keychain)
python3 scripts/keychain.py set ANTHROPIC_API_KEY
# Verify chief-wiggum skills are accessible
# (requires chief-wiggum configured in .claude/settings.local.json)
```
## Usage
```bash
# Recon a target SaaS
/recon trello.com --scope "boards, lists, cards, drag-drop, labels, members, auth"
# Full duplication pipeline (recon → spec → build → compare → loop)
/duplicate trello.com --output plwp/trello-clone --scope "boards, lists, cards"
# Compare clone against target
/compare plwp/trello-clone --target trello.com
# Run gap analysis and feed back into build
/converge plwp/trello-clone --target trello.com
```
Or via the CLI directly:
```bash
duplicat-rex recon trello.com --scope "boards, lists, cards"
duplicat-rex duplicate trello.com --output plwp/trello-clone
```
## Core Loop
```mermaid
graph TD
R[Recon
Multi-source intelligence] --> S[Spec Synthesis
Typed specs with provenance]
S --> T[Test Generation
Dual-execution test cases]
T --> B[Build
via chief-wiggum]
B --> C[Compare
Behavioral conformance]
C --> G{Gap
Analysis}
G -->|gaps found| R
G -->|parity achieved| D[Deploy]
style R fill:#1D76DB,color:#fff
style S fill:#5319E7,color:#fff
style C fill:#D93F0B,color:#fff
style G fill:#FBCA04,color:#000
style D fill:#0E8A16,color:#fff
```
### Recon Sources (ranked by authority)
```mermaid
graph LR
subgraph Authoritative
LA[Live App
Browser-use + Playwright]
AD[API Docs
Official documentation]
end
subgraph Observational
HC[Help Center]
TV[Training Videos
whisper + ffmpeg]
end
subgraph Anecdotal
MK[Marketing Pages]
RD[Reddit / Community]
CL[Changelog]
end
LA & AD --> SS[Spec Store
Facts with provenance]
HC & TV --> SS
MK & RD & CL --> SS
style LA fill:#0E8A16,color:#fff
style AD fill:#0E8A16,color:#fff
style HC fill:#1D76DB,color:#fff
style TV fill:#1D76DB,color:#fff
style MK fill:#D93F0B,color:#fff
style RD fill:#D93F0B,color:#fff
style CL fill:#D93F0B,color:#fff
```
### Fact Lifecycle
```mermaid
stateDiagram-v2
[*] --> UNVERIFIED: create()
UNVERIFIED --> VERIFIED: verify(corroborating_fact_id)
UNVERIFIED --> CONTRADICTED: contradict(contradicting_fact_id)
VERIFIED --> CONTRADICTED: contradict(contradicting_fact_id)
CONTRADICTED --> [*]: supersede(new_fact_id)
```
See [ARCHITECTURE.md](ARCHITECTURE.md) for full details.
## Development
```bash
# Run tests
pytest
# Lint
ruff check .
# Type check
mypy scripts/
```
## Repo Layout
```
.claude/commands/ # Skills: /recon, /duplicate, /compare, /converge
scripts/
├── recon/ # Recon modules (browser, API docs, videos, community)
├── cli.py # Typer CLI entry point
├── keychain.py # Secret management (system keyring)
├── models.py # Canonical data model (Fact, SpecBundle, enums)
├── scope.py # Scope parsing + dependency graph
├── spec_store.py # File-backed spec store with provenance
├── spec_synthesizer.py # LLM synthesis with provenance (planned)
├── compare.py # Behavioral conformance testing (planned)
└── gap_analyzer.py # Gap identification + circuit breaker (planned)
templates/
├── spec-schema.json # JSON Schema Draft 2020-12 for all entities
└── ... # Prompt templates, report formats
tests/ # 224+ pytest tests
```