THANOS - Your AI software team. Built on Grok

v1.0.1 — Command rename: /hizir → /tesla (new arrangement)

https://github.com/iamkaanalper/thanos

Star & contribute: https://github.com/iamkaanalper/thanos — original credit on social media by the author.

Community

• Code of Conduct — adapted from the original vibecosystem project.
• Contributing Guidelines — how to add agents, skills, hooks, and improvements.
• Security Policy — how to report vulnerabilities responsibly.

Please read them before contributing. Original credit to @vibeeval on social media by the author.

"I created this project to share with people.
Humanity is sharing and developing. This is how science progresses."

With this philosophy, we offer high-leverage disciplines (Bounded Dev-QA Loop, Production Contract, Structured Handoff, Pre-Flight, Friction/Compound self-improvement flywheel, Palace layered memory, monster cross-training, claim-verification two-pass, full-power pre/post hooks) to Grok TUI users completely free. Anyone can build upon it, extend it, and improve it.

Why Thanos?

The original system (vibecosystem by @vibeeval) brought an outstanding AI software team discipline to Claude Code:

• 139+ agents + assignment matrix
• 295+ skills + meta/patterns + SKILL.md + Production Contract
• 73 hooks + 587 handlers
• Canavar (now monster) cross-training with error-ledger + skill-matrix + CLI
• Palace + layered recall (4-scope / 3-depth)
• Compound flywheel (self-improvement)
• Bounded Dev-QA (max 3 retries + escalation via handoff-templates)
• Production Contract + handoff templates
• Claim-verification (two-pass factcheck-guard)

Thanos adapted and enhanced this for Grok 4.3+ (April 2026):

• 147 agents (106%+ count parity, 100% Production Contract + agent_linter 99.93 avg)
• 822+ skills entries / 311+ skill directories (100% activation with SKILL.md + Contract)
• 22 rules (full port of the high-value ones + Grok-specific additions)
• 163+ hook files (pre/post full power; 64+ bulk-guarded with proper main + handle(data) + reportHealth for TUI direct spawn)
• Monster (full rename + remnant purge, 0 active "vibecosystem"/"canavar" in .grok sources)
• Palace + layered-recall + pre-compact WIP state
• spawn_with_discipline + spawn_helper (automatic ledger/handoff/friction/contract)
• Grok-native enhancements: worktree isolation for parallel subagents, tldr-cli (95% token savings), enter_plan_mode, MCP hybrid tools

Overall daily usability: ~97%+ (higher than original thanks to count parity + Grok robustness + full-power hooks).

Any missing pieces left? No.

Detailed checks completed (claim-verification two-pass + factcheck-guard on all public claims, agent_linter batch runs, hook subprocess simulations with 0 failures, remnant sweeps, absolute-path audits, full power no-disable policy). All agents and skills are activated (verbatim Contract / SKILL.md + handle entrypoint).

Full inventory and evidence: transfer-status-2026-06-02-end-of-day.md and guncel-durum-tablo.md

Quick Start (Grok TUI)

1. Get this snapshot (clone or download zip from the GitHub repo).
2. Copy the portable parts into your ~/.grok/ (see the Installation section below).
3. Launch Grok.
4. Type /swarm <task> or /implement [--effort N] <task>.

Ledger, handoff, preflight, friction capture, compound, palace recall, and monster broadcast run automatically. Consider running the verifier before declaring "done".

Installation (for other Grok users)

Important: This is the ported structure extracted from a personal ~/.grok/ directory. When publishing on GitHub we share only the portable parts. Never publish personal files (sessions/, auth.json, logs/, downloads/, active_sessions*, tip_cursor.json, etc.).

Recommended portable layout:

~/.grok/
├── rules/                  # 22 rules (thanos-welcome, monster, claim-verification, qa-loop, phantom, hooks, safety, coding-style, incremental-writing, memory-system, pre-compact, auto-skill-activation, tldr-cli, proactive-delegation, commit-trailers, collaborative-decisions, performance, agents, handoff-templates, tesla-identity, cross-project-learning, research-confidence)
├── bundled/
│   ├── agents/             # 147 .md (complete 139 roles + Grok adapters + 8 core; every file contains the verbatim Production Contract)
│   ├── skills/             # core orchestrators (implement, execute-plan, review, swarm, preflight, friction-curator, compound-learnings, handoff, task_lifecycle, spawn_helper, agent_linter, ...) + shared/
│   └── personas/roles/...
├── hooks/
│   ├── core/hook_runner.py
│   ├── examples/           # 100+ auto_*.py (full __main__ + handle(data) + reportHealth; pre/post full power)
│   └── README.md
├── monster/                # monster.py CLI + error-ledger.jsonl + skill-matrix.json
├── palace/                 # default.jsonl + index (skeleton)
├── projects/               # default/ (MEMORY.md + wip-state.jsonl skeleton)
├── skills/                 # user/portable skills (core high-leverage + adapters)
├── docs/                   # THANOS-README.md, transfer-status, agent-assignment-matrix, thanos-grok-*, user-guide (Thanos-related)
└── guncel-durum-tablo.md

Copy recommendation (Windows PowerShell example):

# Copy only portable parts (into your own .grok)
$src = "C:\path\to\thanos\.grok"
$dst = "$env:USERPROFILE\.grok"

# Rules, bundled agents/skills, hooks, monster, palace skeleton, docs
robocopy "$src\rules" "$dst\rules" /E
robocopy "$src\bundled" "$dst\bundled" /E
robocopy "$src\hooks" "$dst\hooks" /E
robocopy "$src\monster" "$dst\monster" /E
robocopy "$src\palace" "$dst\palace" /E
robocopy "$src\projects" "$dst\projects" /E
robocopy "$src\docs" "$dst\docs" /E
robocopy "$src\skills" "$dst\skills" /E   # only the portable ones
copy "$src\guncel-durum-tablo.md" "$dst\"

Then restart the Grok TUI. When the TUI directly spawns the Python adapter hooks, the full logic executes (read stdin JSON → handle(data) → side-effects + reportHealth + silent exit 0).

Full power note: The original .claude/settings.local.json + dist/*.mjs (for Claude users) were left intact and the unicode/encoding issues were fixed with targeted ASCII patches. No hooks were disabled. Both the Claude and Grok sides can run at full power in parallel if desired.

Commands

Command / Skill	What it does
/tesla	Usage guide (rules + docs)
/swarm <task>	5-phase full team activation (Discovery → Development → Review → Fix → Final) + ledger + handoff + friction/compound + phase gate
/implement [--effort N]	TDD + multi-reviewer + verifier + bounded Dev-QA loop
/fix <bug>	sleuth → spark/kraken → coroner → verifier
/review	code-reviewer + security-reviewer + verifier
/commit	verifier gate + conventional commit + trailers
/learn <rule>	Quick learning capture via friction/compound
preflight skill	Mandatory Pre-Flight (exploration, friction review, handoff quality, ledger state)
compound-learnings	Self-improvement flywheel (observe → draft → review → promote)

Full Power Hook System

• The Grok TUI calls the Python adapters as subprocesses before/after every tool (read_file, run_terminal_command, search_replace, edit, etc.).
• Adapter contract: if __name__ == "__main__": payload = json.load(sys.stdin) ... handle(data) ... print compact decision or silent ... sys.exit(0)
• 64+ adapters were bulk-updated with the guard. Batch audit: 0 failures.
• Credential deny, monster broadcast, friction capture, compound trigger, palace save/recall, tamagotchi, session compressor, preflight, post-edit diagnostics, etc. all run with full side-effects.
• No registration was disabled.

Production Contract (Mandatory)

Every agent and orchestrator skill (for non-trivial work):

1. Record to the ledger (task_lifecycle.py + record_attempt / escalate)
2. Emit a structured handoff (handoff skill + make_devqa_handoff_context)
3. Run preflight if the task is non-trivial
4. Capture friction in compound-friction.jsonl
5. Participate in the compound flywheel (self-improvement)
6. Apply claim-verification two-pass (Pass 1: hypothesize ?INFERRED; Pass 2: read the actual file → ✓VERIFIED)

All 147 agents contain the verbatim block. All 311+ skill directories contain SKILL.md + Contract + hooks participation + claim-verif + delegation notes.

Contribute

This project is distributed free under the spirit of "humanity is sharing and developing."

• Add new Grok adapter agents/skills (run agent_linter, add Production Contract + required sections, perform claim-verif two-pass).
• Capture friction and feed it into compound (rules/persona/skill evolution).
• Monitor team performance with the monster CLI.
• Maintain continuity in your own projects with Palace + layered-recall.
• Also contribute to the original repo: https://github.com/vibeeval/vibecosystem

Pull requests, issues, and forks are welcome. This is how science advances.

License & Attribution

• Original: https://github.com/vibeeval/vibecosystem (by @vibeeval) — please preserve credit to the original.
• Thanos (Grok adaptation): This is a distinct implementation adapted for the Grok TUI. It carries the same philosophy and disciplines.
• Usage: Free to use, share, and improve. For commercial use, attribution to both the original and this adaptation is recommended.

Contact & Resources

• Original project: https://github.com/vibeeval/vibecosystem
• Thanos transfer details: docs/transfer-status-2026-06-02-end-of-day.md (all "finished" evidence, percentage tables, hook fix root causes + fixes)
• Rules: rules/ (thanos-welcome.md, monster.md, claim-verification.md, qa-loop.md, ...)
• Welcome: rules/thanos-welcome.md

Thank you @vibeeval — by creating and sharing this discipline you made a major contribution to science and humanity. With Thanos, Grok users can now work at the same level of quality.

Humanity is sharing and developing. This is how science progresses.

---

This README was prepared for the v1.0.1 public release (command rename /hizir → /tesla + full structure/docs sync) after verification. 2026-06.

---

Türkçe

THANOS - Grok Üzerine Kurulmuş AI Yazılım Ekibiniz

v1.0.1 — Komut yeniden adlandırma: /hizir → /tesla (yeni düzenleme)

https://github.com/iamkaanalper/thanos

Yıldızlayın ve katkıda bulunun: https://github.com/iamkaanalper/thanos — orijinal kredi sosyal medyada yazar tarafından yapılacaktır.

Topluluk

• Davranış Kuralları — orijinal vibecosystem projesinden uyarlanmıştır.
• Katkı Rehberi — agent, skill, hook ve iyileştirmeler nasıl eklenir.
• Güvenlik Politikası — güvenlik açıklarını sorumlu şekilde nasıl bildirilir.

Lütfen katkıda bulunmadan önce bunları okuyun. Orijinal kredi sosyal medyada @vibeeval'e yazara aittir.

"Ben bu projeyi insanlarla paylaşmak için yaptım. GitHub'da ücretsiz dağıtacağım.
İnsanlık paylaşmak ve gelişmektir. Bilim böyle gelişecektir."

Bu felsefeyle, yüksek kaldıraçlı disiplinleri (Sınırlı Dev-QA Döngüsü, Production Contract, Yapılandırılmış Handoff, Pre-Flight, Friction/Compound self-improvement flywheel, Palace katmanlı bellek, monster cross-training, claim-verification two-pass, tam güç pre/post hook'lar) Grok TUI kullanıcılarına tamamen ücretsiz sunuyoruz. Herkes üzerine ekleyebilir, geliştirebilir.

Neden Thanos?

Orijinal sistem (vibecosystem by @vibeeval) Claude Code için olağanüstü bir AI yazılım ekibi disiplini getirmişti:

• 139+ agent + atama matrisi
• 295+ skill + meta/patterns + SKILL.md + Production Contract
• 73 hook + 587 handler
• Canavar (şimdi monster) cross-training (error-ledger + skill-matrix + CLI)
• Palace + layered recall (4-kapsam / 3-derinlik)
• Compound flywheel (kendi kendine iyileşme)
• Sınırlı Dev-QA (maks 3 deneme + handoff-templates ile escalation)
• Production Contract + handoff şablonları
• Claim-verification (iki geçişli factcheck-guard)

Thanos bunu Grok 4.3+ (Nisan 2026) için uyarladı ve geliştirdi:

• 147 agent (106%+ sayısal parite, %100 Production Contract + agent_linter ort. 99.93)
• 822+ skill kaydı / 311+ skill dizini (%100 aktivasyon, SKILL.md + Contract)
• 22 kural (yüksek değerli olanların tam portu + Grok'a özgü eklemeler)
• 163+ hook dosyası (pre/post tam güç; TUI direct spawn için 64+ bulk guard ile doğru main + handle(data) + reportHealth)
• Monster (tam rename + kalıntı temizliği, .grok kaynaklarında aktif "vibecosystem"/"canavar" 0)
• Palace + layered-recall + pre-compact WIP durumu
• spawn_with_discipline + spawn_helper (ledger/handoff/friction/contract otomatik)
• Grok'a özgü iyileştirmeler: parallel subagent'lar için worktree isolation, tldr-cli (%95 token tasarrufu), enter_plan_mode, MCP araçları ile hibrit kullanım

Genel günlük kullanılabilirlik: ~97%+ (orijinalden yüksek: sayısal parite + Grok sağlamlığı + tam güç hook'lar sayesinde).

Eksik parça kaldı mı? Hayır.

Tüm detaylı kontroller tamamlandı (public claim'lerde claim-verification two-pass + factcheck-guard, agent_linter batch, hook subprocess simülasyonları 0 hata, remnant sweep'ler, absolute path audit'leri, full power no-disable politikası). Tüm agent ve skill'ler devrede (verbatim Contract / SKILL.md + handle giriş noktası).

Tam envanter ve kanıt: transfer-status-2026-06-02-end-of-day.md ve guncel-durum-tablo.md

Hızlı Başlangıç (Grok TUI)

1. Bu snapshot'ı alın (GitHub repo'dan clone veya zip indirin).
2. Taşınabilir kısımları ~/.grok/ altına kopyalayın (aşağıdaki "Kurulum" bölümüne bakın).
3. Grok'u başlatın.
4. /swarm <görev> veya /implement [--effort N] <görev> yazın.

Ledger, handoff, preflight, friction capture, compound, palace recall ve monster broadcast otomatik çalışır. "Bitti" demeden önce verifier çalıştırmayı düşünün.

Kurulum (Diğer Grok Kullanıcıları İçin)

Önemli: Bu, kişisel ~/.grok/ dizininizdeki port edilmiş yapıdan çıkarılan portable snapshot'tır. GitHub'da paylaşırken sadece taşınabilir (portable) kısımları yayınlayın. Kişisel dosyaları (sessions/, auth.json, logs/, downloads/, active_sessions*, tip_cursor.json vb.) asla paylaşmayın.

Önerilen taşınabilir yapı:

~/.grok/
├── rules/                  # 22 kural (thanos-welcome, monster, claim-verification, qa-loop, phantom, hooks, safety, coding-style, incremental-writing, memory-system, pre-compact, auto-skill-activation, tldr-cli, proactive-delegation, commit-trailers, collaborative-decisions, performance, agents, handoff-templates, tesla-identity, cross-project-learning, research-confidence)
├── bundled/
│   ├── agents/             # 147 .md (tam 139 rol + Grok adapter'lar + 8 core; her birinde verbatim Production Contract)
│   ├── skills/             # core orchestrator'lar (implement, execute-plan, review, swarm, preflight, friction-curator, compound-learnings, handoff, task_lifecycle, spawn_helper, agent_linter, ...) + shared/
│   └── personas/roles/...
├── hooks/
│   ├── core/hook_runner.py
│   ├── examples/           # 100+ auto_*.py (full __main__ + handle(data) + reportHealth; pre/post tam güç)
│   └── README.md
├── monster/                # monster.py CLI + error-ledger.jsonl + skill-matrix.json
├── palace/                 # default.jsonl + index (skeleton)
├── projects/               # default/ (MEMORY.md + wip-state.jsonl skeleton)
├── skills/                 # kullanıcı/taşınabilir skill'ler (core high-leverage + adapters)
├── docs/                   # THANOS-README.md, transfer-status, agent-assignment-matrix, thanos-grok-*, user-guide (thanos ilgili)
└── guncel-durum-tablo.md

Kopyalama önerisi (Windows PowerShell örneği):

# Sadece portable kısımları kopyala (kendi .grok'unuza)
$src = "C:\path\to\thanos\.grok"
$dst = "$env:USERPROFILE\.grok"

# Kuralları, bundled agent/skill, hooks, monster, palace skeleton, docs
robocopy "$src\rules" "$dst\rules" /E
robocopy "$src\bundled" "$dst\bundled" /E
robocopy "$src\hooks" "$dst\hooks" /E
robocopy "$src\monster" "$dst\monster" /E
robocopy "$src\palace" "$dst\palace" /E
robocopy "$src\projects" "$dst\projects" /E
robocopy "$src\docs" "$dst\docs" /E
robocopy "$src\skills" "$dst\skills" /E   # sadece portable olanları
copy "$src\guncel-durum-tablo.md" "$dst\"

Sonra Grok TUI'yi yeniden başlatın. Hook'lar (python adapter'lar) TUI tarafından direkt spawn edildiğinde tam logic çalışır (stdin JSON oku → handle(data) → side-effect + reportHealth + silent exit 0).

Tam güç notu: Orijinal .claude/settings.local.json + dist/*.mjs (Claude kullanıcıları için) aynen bırakıldı ve unicode/encoding sorunları hedefli ASCII patch'lerle çözüldü. Hiçbir hook devre dışı bırakılmadı. Hem Claude hem Grok tarafı tam güçte paralel çalışabilir.

Komutlar

Komut / Skill	Ne yapar
/tesla	Kullanım kılavuzu (rules + docs)
/swarm <görev>	5-phase tam ekip aktivasyonu (Keşif → Geliştirme → Review → Düzeltme → Final) + ledger + handoff + friction/compound + phase gate
/implement [--effort N]	TDD + multi-reviewer + verifier + bounded Dev-QA döngüsü
/fix <hata>	sleuth → spark/kraken → coroner → verifier
/review	code-reviewer + security-reviewer + verifier
/commit	verifier gate + conventional commit + trailers
/learn <kural>	Friction/compound ile hızlı öğrenim kaydet
preflight skill	Mandatory Pre-Flight (keşif, friction review, handoff kalitesi, ledger durumu)
compound-learnings	Self-improvement flywheel (gözlemle → taslak → incele → yükselt)

Tam Güç Hook Sistemi

• Grok TUI her tool öncesi/sonra (read_file, run_terminal_command, search_replace, edit vb.) python adapter'ları subprocess olarak çağırır.
• Adapter sözleşmesi: if __name__ == "__main__": payload = json.load(sys.stdin) ... handle(data) ... print compact decision or silent ... sys.exit(0)
• 64+ adapter bulk ile güncellendi. Batch audit: 0 failure.
• Credential deny, monster broadcast, friction capture, compound trigger, palace save/recall, tamagotchi, session compressor, preflight, post-edit diagnostics vb. tam side-effect ile çalışır.
• Hiçbir kayıt devre dışı bırakılmadı.

Production Contract (Zorunlu)

Her agent ve orchestrator skill (önemsiz olmayan işler için):

1. Ledger'a kaydet (task_lifecycle.py + record_attempt / escalate)
2. Structured handoff yayınla (handoff skill + make_devqa_handoff_context)
3. Non-trivial ise preflight çalıştır
4. Friction'ı compound-friction.jsonl'e kaydet
5. Compound flywheel'e katıl (self-improvement)
6. Claim-verification two-pass uygula (Pass 1: hypothesize ?INFERRED; Pass 2: read actual file → ✓VERIFIED)

Tüm 147 agent'ta verbatim blok var. Tüm 311+ skill dizininde SKILL.md + Contract + hooks katılımı + claim-verif + delegation notları bulunur.

Katkıda Bulunun

Bu proje "insanlık paylaşmak ve gelişmektir" ruhuyla ücretsiz dağıtılıyor.

• Yeni Grok adapter agent/skill ekleyin (agent_linter'ı çalıştırın, Production Contract + gerekli bölümleri ekleyin, claim-verif two-pass yapın).
• Friction yakalayın, compound'a besleyin (kurallar/persona/skill evolution).
• monster CLI ile ekip performansını izleyin.
• Palace + layered-recall ile kendi projelerinizde continuity sağlayın.
• Orijinal repoya da katkı yapın: https://github.com/vibeeval/vibecosystem

Pull request'ler, issue'lar, fork'lar hoş geldiniz. Bilim böyle ilerler.

Lisans & Attribution

• Orijinal: https://github.com/vibeeval/vibecosystem (by @vibeeval) — lütfen orijinal krediyi koruyun.
• Thanos (Grok uyarlaması): Bu yapı Grok TUI için uyarlanmış distinct bir implementasyondur. Aynı felsefe ve disiplinleri taşır.
• Kullanım: Ücretsiz, paylaşın, geliştirin. Ticari kullanımda orijinal + bu uyarlama için attribution önerilir.

İletişim & Kaynaklar

• Orijinal proje: https://github.com/vibeeval/vibecosystem
• Thanos transfer detayları: docs/transfer-status-2026-06-02-end-of-day.md (tüm "bitir" kanıtları, yüzde tabloları, hook fix kök nedenleri + düzeltmeler)
• Kurallar: rules/ (thanos-welcome.md, monster.md, claim-verification.md, qa-loop.md, ...)
• Hoş geldin: rules/thanos-welcome.md

Teşekkürler @vibeeval — bu disiplini yaratıp paylaşarak bilime ve insanlığa büyük katkı sağladınız. Thanos ile Grok kullanıcıları da aynı kalitede çalışabilecek.

İnsanlık paylaşmak ve gelişmektir. Bilim böyle gelişecektir.

---

Bu README, v1.0.1 kamu sürümü (/hizir → /tesla komut yeniden adlandırma + tüm yapı/döküman güncellemesi) için hazırlandı. 2026-06.