Home

8-Habit AI Dev

Important

Anti-Vibe-Coding plugin for Claude Code. A 7-step workflow discipline and 8-Habit cross-verification framework that replaces ad-hoc AI prompting with structured, reviewable, governable development.

"ทำเสร็จ ≠ ทำดี" — Done is not Done well.

Why This Exists

AI-assisted coding is fast but often undisciplined: missing requirements, skipped reviews, no rollback plans, accumulating tech debt. This plugin enforces the habits that separate effective developers from vibe coders — based on Stephen Covey's 7 Habits + The 8th Habit, adapted for the AI era.

At a Glance

23 skills	8 workflow + 5 assessment + 3 meta + 2 debug + 2 compliance + 3 helpers (orchestrator/spec/audience)
8 Habits	Covey's 7 + The 8th, adapted for AI-assisted development
17 ADRs	Every non-trivial decision documented
4 maturity levels	Dependence → Independence → Interdependence → Significance
Zero dependencies	Pure markdown — no npm, no build step
EU AI Act ready	First Claude Code plugin with compliance toolkit
External review	Scored 9.5/10 (EXCELLENT) on first formal audit

Tip

New here? Start with Installation then Getting Started. Already installed? Jump to Workflow Overview or type /using-8-habits in Claude Code.

The 7-Step Workflow

Legend:  [O] optional   [H] human checkpoint   [!] NEVER SKIP

   [O] 0 · /research        H5 Understand    investigate before specifying
        │
        ▼
   [H] 1 · /requirements    H2 End in Mind   define what, why, who
        │
        ▼
   [H] 2 · /design          H8 Voice         architecture (human-led)
        │
        ▼
   [O] 3 · /breakdown       H3 First Things  atomic tasks, no scope creep
        │
        ▼
   [O] 4 · /build-brief     H5 Understand    context before coding
        │
        ▼
   [!] 5 · /review-ai       H4 Win-Win       audit before commit
        │
        ▼
   [H] 6 · /deploy-guide    H1 Proactive     staging first, rollback ready
        │
        ▼
   [O] 7 · /monitor-setup   H7 Sharpen Saw   observe after deploy

Step	Command	Habit	Purpose
0	/research	H5	Investigate before specifying
1	/requirements	H2	Define what, why, who
2	/design	H8	Architecture decisions (human-led)
3	/breakdown	H3	Atomic tasks, no scope creep
4	/build-brief	H5	Context before coding
5	/review-ai	H4	Audit before commit
6	/deploy-guide	H1	Staging first, rollback ready
7	/monitor-setup	H7	Observe after deploy

Beyond the Workflow

Assessment — run at any point for deeper analysis:

Skill	Purpose
/cross-verify	17-question 8-Habit checklist
/whole-person-check	Body/Mind/Heart/Spirit balance
/security-check	OWASP Top 10 focused review
/scrutinize	Outsider-perspective review — questions if change should exist
/consistency-check	Cross-artifact drift detection (PRD ↔ design ↔ tasks)
/reflect	Post-task retrospective with lesson persistence
/workflow	Interactive guided walkthrough

Debug Discipline — active investigation + post-fix record:

Skill	Purpose
/diagnose	6-phase active bug investigation (feedback-loop first)
/post-mortem	Canonical RCA writeup after a validated fix lands

Spec & Communication — helpers orthogonal to the 7-step workflow:

Skill	Purpose
/save-spec	Scaffold project-root SPEC.md orientation hub
/management-talk	Reshape engineer content for leadership channels

Meta & Onboarding — learn and adapt:

Skill	Purpose
/using-8-habits	Onboarding decision tree — which skill when?
/calibrate	Self-assess maturity, adapt guidance verbosity

Compliance & Audit — governance and transparency:

Skill	Purpose
/eu-ai-act-check	EU AI Act 9-obligation checklist (Articles 9-15)
/ai-dev-log	AI-assisted development log from git history

Principles

Note

Skills are read-only guidance — they tell Claude how to approach a task. They never edit files themselves.

• Human-in-the-loop for architecture and irreversible decisions
• Zero dependencies: pure markdown + bash validation
• PR-first: every change goes through review, even documentation
• Complementary: pairs with claude-governance for enforcement + compliance

Reference

• 8 Habits — the full playbook
• Skills Catalog — all 23 skills, when to use each
• Architecture — how the plugin works internally
• Maturity Model — adaptive guidance levels
• Vibe Coding vs Structured — side-by-side comparison
• FAQ · Troubleshooting

Contributing

See Contributing to Wiki. The wiki is a build artifact — edit docs/wiki/<page>.md and open a PR.