รุ่นปัจจุบัน: v2.0.0 — Knowledge-Graph-Assisted Architecture Passes
Cognitive framework ที่เปลี่ยน AI agents ให้เป็น senior architects — evidence-first system understanding, honest uncertainty mapping, architecture debt and convention assessment, knowledge-graph-assisted dependency mapping, impact and blast radius analysis, และ architecture reasoning ก่อนลงมือ
Skill นี้บังคับให้ agent inspect, map, assess, document, validate, และรายงานระบบที่มีอยู่จริง ก่อนเสนอการเปลี่ยนแปลงสถาปัตยกรรม โดยเริ่มจาก smallest safe architecture pass เสมอ
Intake → Inspect → Classify → Question → Map → Assess → Document → Validate → Reportเลือก smallest safe pass:
| Mode | เมื่อไหร่ | Output |
|---|---|---|
| Scan | งาน bounded, exploratory, low-risk | 1 compact architecture note |
| Focus | 1 module, workflow, หรือ boundary ที่ชัดเจน | 1-3 artifacts |
| Full | Whole-system, handoff, 3+ modules interacting, persistence, security, integrations | Full output package |
เริ่มที่ Scan เสมอ — เลื่อนขึ้นเมื่อ scope, evidence, risk, หรือ handoff needs ต้องการ โดยต้องมี trigger + evidence + risk statement
From powerful models to elite senior architects: skill เพิ่มชั้น cognitive framework รอบโมเดลที่มีความสามารถ
Full workflow infographic: core flow, mode selection, evidence gates, repository anatomy, และ expected outputs
Core flow animation: inspect → classify → question → map → assess → document → validate → report
| ใช้ ✅ | ไม่ใช้ ❌ |
|---|---|
| Agent ต้อง map existing codebase หรือระบบซอฟต์แวร์ | งานเป็นไอเดียดิบล้วนๆ — ใช้ $idea-to-architecture-agent แทน |
| ต้องการหา architecture debt, convention drift, boundary violations, flow conflicts | แค่ implementation fix เล็กๆ ที่ไม่มีผลต่อสถาปัตยกรรม |
| ทีมต้องการ module maps, Mermaid diagrams, handoff notes | Output จะกลายเป็น decorative documentation |
| review การเปลี่ยนแปลงสถาปัตยกรรมก่อน approve |
- Inspect ระบบจริงก่อน — file search, git history, Mermaid checks, และ codebase graph queries (เมื่อมี MCP)
- Map dependencies และ boundaries — optional knowledge-graph-assisted: graph data เร่งหา dependency chains, blast radius, circular dependencies โดย map เข้า evidence taxonomy เดิม (
Direct/Inferred) — file reading ชนะเมื่อ graph ขัดแย้ง - Assess — ชี้ debt, drift, boundary violations, flow conflicts ทุก finding ต้องมี evidence, impact, severity, confidence, smallest safe correction
- Validate — 3 gates: claim traceability, scope alignment, handoff readiness
- Output — architecture overview, module map, data flow, workflow map, debt register, risk register, open questions, ai-agent notes, Mermaid diagrams
| Skill | ใช้เมื่อ |
|---|---|
senior-architect-agent (นี้) |
ระบบที่มีอยู่, existing + proposed changes, architecture review, handoff |
$idea-to-architecture-agent |
ไอเดียดิบ, product concept, feature request, เป้าหมายธุรกิจที่ยังไม่มี implementation |
- ตัด Idea-to-Architecture Mode → route ไป
$idea-to-architecture-agent - เพิ่ม Assess step —
rules/assessment-rules.md,templates/debt-register.md - Knowledge-graph-assisted mapping และ assessment (optional) — graph cycles, fan-in outliers, layer violations → signal ป้อน dimension เดิม, ไม่ใช่ finding category ใหม่
- Rule 0 — default Scan Mode, promotion ต้องมี trigger + evidence + risk
- 7 rules files, 11 templates, example output
ดูทั้งหมดใน CHANGELOG.md
examples/basic-web-app/output/
├── architecture-overview.md
├── system-boundary.md
├── module-map.md
├── data-flow.md
├── workflow-map.md
├── file-responsibility-map.md
├── debt-register.md
├── risk-register.md
├── diagram.mmd / diagram.svg
├── open-questions.md
└── ai-agent-notes.mdsenior-architect-agent/
├── README.md / SKILL.md / INSTALL.md / CHANGELOG.md / LICENSE
├── assets/visuals/ # infographic, workflow GIF, Mermaid source
├── scripts/ # build tooling (build_core_flow_gif.py)
├── agents/ # interface metadata (openai.yaml)
├── adapters/ # AGENTS.example.md
├── docs/ # philosophy, workflow, output-spec, diagram-guidelines, model-requirements, anti-patterns
├── rules/ # 7 rule files (inspection, question, assessment, documentation, diagram, anti-overengineering, agent-handoff)
├── templates/ # 11 output templates
└── examples/ # basic-web-app output
ต้องการ long context (128K+ ปกติ, 200K+ สำหรับ multi-service) และ instruction following ที่แข็งแรง 32K พอสำหรับ Scan Mode
ดู INSTALL.md สำหรับ Codex, Claude Code, Antigravity, AGENTS.md, และ manual
aetox-skills/senior-architect-agentMIT