Skip to content
Back to skills

Project Audit

ASecurity

Use this skill when asked to audit, assess, or health-check an existing software project or codebase - "audit project ini", "kondisi codebase gimana", "layak dilanjut/diambil alih?", due diligence on inherited code, "kenapa project ini lambat terus", or producing a findings report with priorities. Produces a scored, evidence-based audit report. Do NOT use for reviewing a single diff/PR (code-review-hardening) or a live incident (incident-debugging).

  • 2 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added October 15, 2026
ai-agentstestingdebuggingcode-reviewgit

Security analysis

A100/100

Scanned October 15, 2026

npx -y skills add me-workspace/orchestrator-operational-agent-skills --skill project-audit --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Project Audit?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Project Audit
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/me-workspace-project-audit/badge)](https://www.skillsdirectory.com/skills/me-workspace-project-audit)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: project-audit
description: Use this skill when asked to audit, assess, or health-check an existing software project or codebase - "audit project ini", "kondisi codebase gimana", "layak dilanjut/diambil alih?", due diligence on inherited code, "kenapa project ini lambat terus", or producing a findings report with priorities. Produces a scored, evidence-based audit report. Do NOT use for reviewing a single diff/PR (code-review-hardening) or a live incident (incident-debugging).
---

# Project Audit

An audit is evidence + score + prioritized fixes - never impressions. Every finding cites a file/config/commit; every recommendation has effort and impact.

## Audit dimensions (score each 1-5, with written evidence)

**1. Arsitektur & kode**
Struktur modul jelas dan konsisten? Dependency antar modul searah atau spaghetti? Duplikasi besar (cek dengan grep pola berulang)? Dead code? Ukuran file/fungsi ekstrem? Framework/library versi berapa jauh di belakang?

**2. Keamanan** (pakai lensa code-review-hardening pada skala repo)
Secrets di repo/history (`git log -p | grep -i` untuk key/password/token)? Auth di setiap endpoint sensitif? Input validation di boundary? Dependency dengan CVE dikenal? Konfigurasi produksi (CORS, debug mode, verbose errors)?

**3. Data & migrasi**
Skema terdokumentasi? Migrasi versioned dan reversible? Backup ada, off-box, dan pernah di-restore? Data sensitif terenkripsi/di-mask di log?

**4. Testing & CI**
Coverage alur kritis (bukan angka % global - cek spesifik: auth, pembayaran, data-write utama). CI jalan dan hijau? Berapa lama? Tes flaky? Bisakah orang baru menjalankan tes secara lokal dalam < 30 menit?

**5. Operasional**
Deploy: berapa langkah manual? Rollback pernah dicoba? Logging cukup untuk debug insiden nyata? Monitoring/alerting ada? Single point of failure (satu server, satu orang, satu cron)?

**6. Deliverability (kesehatan proses)**
Dari git history: frekuensi commit/merge, ukuran PR, berapa lama PR menggantung, bus factor (berapa % commit dari 1 orang), TODO/FIXME yang membusuk, issue tracker vs realita kode.

**7. Dokumentasi & onboarding**
README bisa membuat orang baru menjalankan project? Keputusan arsitektur tercatat? Runbook operasional ada?

## Method (in order - don't skip the cheap steps)
1. **Recon**: baca README, struktur folder, file konfigurasi, CI config, dependency manifest. 30 menit ini membentuk peta.
2. **Git archaeology**: `git log --stat`, kontributor, hotspot file (file yang paling sering diubah = pusat risiko), umur branch.
3. **Automated sweep**: dependency audit, grep secrets, hitung LOC/duplikasi, jalankan tes.
4. **Deep dive** hanya pada 3-5 area yang recon tunjuk sebagai berisiko - bukan seluruh repo merata.
5. **Verify findings**: tiap temuan besar dicek ulang di kode aslinya sebelum masuk laporan.

## Report format
```
# Audit <project> - <tanggal>
## Ringkasan eksekutif (≤ 5 kalimat: kondisi umum, risiko terbesar, rekomendasi utama)
## Skor per dimensi (tabel 1-5 + satu kalimat bukti per dimensi)
## Temuan (urut prioritas)
[P1-BLOCKER|P2-HIGH|P3-MEDIUM|P4-LOW] <judul>
  Bukti: <file:line / commit / output perintah>
  Dampak: <apa yang rusak/berisiko, dalam istilah bisnis bila mungkin>
  Rekomendasi: <aksi konkret> | Effort: S/M/L
## Roadmap perbaikan: batch 1 (blocker), batch 2 (high), debt register sisanya
## Yang SUDAH baik (wajib - kalibrasi kepercayaan laporan)
```

## Verdict scale (for "layak dilanjut?" questions)
**Sehat** (lanjut normal) · **Perlu perawatan** (alokasikan 20-30% kapasitas ke perbaikan) · **Kritis** (hentikan fitur baru, stabilisasi dulu) · **Jangan diambil alih / tulis ulang bertahap** - sertakan estimasi biaya tiap jalur. Jangan pernah memberi verdict tanpa menyebut apa yang akan mengubahnya.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.