# ARCHITECTURE.md — Labs Pentest

> Hitungan nyata repo (19 Nov 2026, diverifikasi dari tree): **14 mode**,
> **25 skrip Python**, **4 skrip shell**, 3 dokumen inti
> (`AGENTS.md`, `MODES.md`, `PLAYBOOK.md`), 1 template laporan,
> 2 file registry target. Diagram visual: `docs/diagrams.html`.

## 1. Peta repo

```text
labs-pentest/
├── AGENTS.md / MODES.md / PLAYBOOK.md   # doktrin: guardrail, menu mode, taktik
├── targets/
│   ├── AUTHORIZED.md                    # registry aset terotorisasi (Tier 0/1/1w/1L)
│   └── SCOPE.md                         # aturan otorisasi + penanganan data
├── scripts/                             # 25 .py + 4 .sh (lihat §2)
│   ├── recon/ api/ osint/               # cari & petakan permukaan serang
│   ├── auth-attack/ evasion/ captcha/   # lawan login, WAF, CAPTCHA
│   ├── blue/                            # sisi bertahan: audit + hardening
│   ├── c2/ cctv/ wifi/ malware/         # OOB, perangkat, sampel
│   ├── new-engagement.sh                # scaffold engagement baru
│   ├── advisory-sweep.py                # OSV sweep dependensi (CI)
│   ├── check-static.mjs                 # gate P0 anti-slop + OG (CI)
│   ├── make-og-image.py                 # og-image.png via PIL
│   └── verify-live.py                   # verifikasi live Pages
├── reports/TEMPLATE.md                  # template laporan 10 bagian
├── tests/test_smoke.py                  # pytest smoke (tanpa target nyata)
├── docs/                                # PRD → arsitektur → pakai → uji → deploy
├── index.html / styles.css / app.js     # landing statis (etalase metodologi)
├── og-image.png                         # preview kartu (1200x630, PIL)
├── requirements.txt                     # 6 dependensi produksi, pin ==
├── package.json                         # tooling landing (html-validate, playwright)
└── .github/workflows/ci.yml             # ruff + shellcheck + pytest + OSV + E2E
```

## 2. Katalog skrip

### recon/ — permukaan serang web (4 skrip, stdlib saja)

| Skrip | Tugas | Dependensi |
|---|---|---|
| `headers_audit.py` | audit header/CSP/HSTS/CORS/cookie | stdlib |
| `exposure.py` | pindai file sensitif terekspos (`.git/HEAD`, `.env`, backup) | stdlib |
| `js_secrets.py` | tambang secret/endpoint dari bundle JS | stdlib |
| `jwt_probe.py` | audit struktur JWT + crack HS256 (wordlist) | stdlib |

### api/ — logika API (5 skrip)

| Skrip | Tugas | Dependensi |
|---|---|---|
| `fuzz.py` | fuzz type-confusion (16 payload) | stdlib |
| `race.py` | uji race condition (N request paralel) | stdlib |
| `graphql_probe.py` | introspeksi + mutasi + batching GraphQL | stdlib |
| `param_miner.py` | mining parameter tersembunyi | stdlib |
| `ws_probe.py` | audit auth/origin WebSocket | `websocket-client` |

### osint/ — jejak pasif (5 skrip + 1 dork pack)

| Skrip | Tugas | Dependensi |
|---|---|---|
| `ct_subs.py` | enum subdomain via Certificate Transparency | stdlib |
| `ct_watch.py` | pantau subdomain baru vs baseline | stdlib |
| `dns_recon.py` | A/MX/TXT/SOA/CAA + DMARC + SPF | `dnspython` |
| `wayback_urls.py` | URL arsip Wayback (endpoint lama) | stdlib |
| `site_map.py` | scraping sopan + umpan `js_secrets` | stdlib |
| `dorks.md` | 7 paket dork (secret, backup, admin, GitHub, Shodan, CCTV) | — |

### auth-attack/, evasion/, captcha/

| Skrip | Tugas | Dependensi |
|---|---|---|
| `auth-attack/spray.py` | password spraying low-and-slow | stdlib |
| `evasion/req.py` | request fingerprint-TLS browser + deteksi CAPTCHA/WAF | `curl-cffi` |
| `evasion/redirector-setup.sh` | setup redirector WireGuard di VPS (butuh root) | shell |
| `captcha/audio_solver.py` | transkrip audio challenge reCAPTCHA | `faster-whisper` (opsional, lazy) |
| `captcha/geetest_gap.py` | deteksi offset-x gap slider GeeTest | `opencv-python-headless`, `numpy` |

### blue/ — sisi bertahan (3 py + 1 sh)

| Skrip | Tugas | Dependensi |
|---|---|---|
| `tls_audit.py` | audit TLS (versi, cipher, expiry, chain) | stdlib |
| `smb_audit.py` | audit SMB (SMBv1, null session, share) | `impacket` |
| `supply_audit.py` | audit supply-chain (dep floating, registry http, workflow) | stdlib |
| `lotl_audit.sh` | inventaris biner Living-off-the-Land | shell |

### c2/, cctv/, wifi/, malware/

| Skrip | Tugas | Dependensi |
|---|---|---|
| `c2/listener.py` | listener OOB (blind SSRF / egress test) | stdlib |
| `cctv/onvif_discover.py` | discovery kamera ONVIF di LAN | stdlib |
| `cctv/rtsp_probe.py` | uji stream terbuka + kredensial default | stdlib |
| `cctv/default_creds.txt` | daftar kredensial default publik (uji ke milik sendiri saja) | — |
| `wifi/wifi_preflight.sh` | cek kesiapan hardware WiFi (tanpa sudo) | shell |
| `malware/triage.py` | triase statis sampel (hash, strings, IOC, PE, macro) — tanpa eksekusi | `pefile`/`oletools` (opsional, lazy) |

## 3. Aliran data engagement

```text
deklarasi / surat izin ──> targets/AUTHORIZED.md (registry)
        │
        ▼
scripts/new-engagement.sh ──> evidence/<tgl>-<nama>/ (raw/ chmod 700)
        │
        ▼
recon → scan → enum → PoC (maks 5 record) ──> data mentah HANYA di raw/
        │
        ▼
reports/TEMPLATE.md ──> laporan MASKED (format + jumlah + hash)
        │
        ▼
blue-team follow-up + retest ──> raw/ dihapus, penghapusan dicatat
```

Aturan keras: data mentah tidak pernah ke chat, laporan, gateway,
atau repo. `.gitignore` memblokir `evidence/` dan `wordlists/`.

## 4. Keputusan arsitektur

- **Stdlib dulu.** 17 dari 25 skrip Python hanya butuh stdlib — sengaja,
  supaya recon dasar jalan di mesin apa pun tanpa install.
- **Lazy import untuk dependensi berat.** `faster-whisper`, `pefile`,
  `oletools`, `impacket`, `websocket-client` diimpor di dalam fungsi /
  dijaga `try/except` dengan pesan install yang jelas — skrip tetap
  memberi `--help` walau dependensinya belum terpasang.
- **Pin exact di requirements.txt.** Semua dependensi produksi `==`,
  supaya advisory-sweep OSV bisa memeriksa versi yang persis dipakai CI.
- **Landing terpisah dari produk.** Landing statis di root hanya etalase
  metodologi; tidak ada skrip playbook yang dijalankan dari browser.
