code-ai-installer 2.3.0 → 2.4.0

package/domains/content/agents/reviewer.md

@@ -43,6 +43,7 @@
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - `$content-review-checklist` — системный чеклист (38 пунктов, 8 категорий)
 - `$brand-compliance` — ToV, visual identity, messaging, юридика
 

package/domains/content/agents/strategist.md

@@ -38,6 +38,7 @@
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - `$audience-analysis` — анализ ЦА: персоны, боли, триггеры, платформенные предпочтения
 - `$platform-strategy` — выбор платформ и форматов, paid/organic, roadmap
 - `$content-brief` — формирование полного брифа для Researcher + Copywriter

package/domains/content/agents/visual_concept.md

@@ -39,6 +39,7 @@
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - `$visual-brief` — детальное ТЗ на визуал
 - `$platform-visual-specs` — размеры, форматы, safe zones
 - `$brand-guidelines` — палитра, шрифты, лого, стиль

package/domains/development/agents/architect.md

@@ -273,6 +273,7 @@
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $current-state-analysis
 - $system-design-checklist
 - $architecture-doc

package/domains/development/agents/conductor.md

@@ -275,6 +275,7 @@
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $board
 - $handoff
 - $memory

package/domains/development/agents/devops.md

@@ -193,6 +193,7 @@ lint → typecheck → unit tests → integration tests → build → deploy (st
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $deployment-ci-plan
 - $docker-kubernetes-architecture
 - $k8s-manifests-conventions

package/domains/development/agents/product_manager.md

@@ -122,6 +122,7 @@ PRD должен содержать:
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $pm-interview
 - $pm-prd
 - $pm-backlog

package/domains/development/agents/reviewer.md

@@ -154,6 +154,7 @@ Reviewer обязан выдать отчёт, который может исп
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $code-review-checklist
 - $security-review
 - $cloud-infrastructure-security

package/domains/development/agents/senior_full_stack.md

@@ -170,6 +170,7 @@
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $tdd-workflow
 - $testing-strategy-js
 - $tests-quality-review

package/domains/development/agents/tester.md

@@ -210,6 +210,7 @@ Tester не обязан писать всю автоматизацию сам,
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $qa-test-plan
 - $qa-manual-run
 - $qa-browser-testing — визуальное E2E через встроенный Antigravity Browser

package/domains/development/agents/ux_ui_designer.md

@@ -153,6 +153,7 @@
 ---
 
 ## Используемые skills (вызовы)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - $ux-discovery
 - $ux-spec
 - $ui-inventory

package/domains/product/agents/conductor.md

@@ -46,6 +46,7 @@ Build-Camp vs Cut-Camp в Full B) провели независимый анал
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **$board** — управление доской задач, трекинг статусов гейтов, метрики прогресса
 - **$handoff** — формирование handoff-конвертов и межсессионных файлов
 - **$gates** — контроль гейтов, проверка deliverable, severity (Blocker/Gap/Note)

package/domains/product/agents/data_analyst.md

@@ -50,6 +50,7 @@ guardrails + leading indicators, (2) каждая гипотеза formalized п
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$saas-metrics`** — B2B SaaS метрики (MRR/ARR, churn, NRR, LTV/CAC, Payback, Rule of 40)
 - **`$hypothesis-template`** — формализация гипотез
 - **`$gates`** — проверка deliverable по DATA-01 criteria

package/domains/product/agents/designer.md

@@ -48,6 +48,7 @@ handoff — достаточно для Layouter (LY-01) для генераци
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$report-design`** — макет PRD / Deck (information architecture, typography, visual hierarchy)
 - **`$gates`** — проверка deliverable по DS-01 criteria
 - **`$handoff`** — формирование handoff-конверта для Layouter

package/domains/product/agents/discovery.md

@@ -47,6 +47,7 @@ output становится общим входом для Camp Alpha и Beta.
 ## Используемые skills
 
 ### Обязательные
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$user-interview-script`** — скрипт B2B-интервью (buyer + end-user треки)
 - **`$jtbd-canvas`** — JTBD канва (functional / emotional / social jobs)
 - **`$problem-statement`** — чёткая формулировка проблемы

package/domains/product/agents/layouter.md

@@ -47,6 +47,7 @@ colors, spacing), (4) accessibility — semantic HTML, proper headings, alt-text
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$html-pdf-report`** — HTML/CSS → PDF pipeline (print-ready, self-contained)
 - **`$gates`** — проверка deliverable по LY-01 criteria
 - **`$handoff`** — формирование handoff-конверта для Release Gate

package/domains/product/agents/mediator.md

@@ -53,6 +53,7 @@ Beta / hybrid, с явным обоснованием для каждого эл
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$handoff`** — приём от COND-04 + формирование конверта с Unified Synthesis
 - **`$gates`** — проверка deliverable по MED-xx criteria перед передачей
 - **`$board`** — обновление статуса MED-01

package/domains/product/agents/pm.md

@@ -54,6 +54,7 @@ metrics (NSM/OKR link), (5) out-of-scope явно (prevent creep), (6) camp disc
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$prd-template`** — Product Requirements Document
 - **`$user-story`** — INVEST user stories
 - **`$acceptance-criteria`** — Gherkin AC

package/domains/product/agents/product_strategist.md

@@ -53,6 +53,7 @@ Reasoning всегда `high` — стратегические решения в
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$product-vision`** — видение продукта (vision statement + principles)
 - **`$okr-framework`** — OKR для квартала/года
 - **`$north-star-metric`** — NSM selection + leading indicators

package/domains/product/agents/tech_lead.md

@@ -52,6 +52,7 @@ critical path и зависимостями, (5) camp discipline (Full B).
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$epic-breakdown`** — декомпозиция initiative на epics → stories с зависимостями
 - **`$gates`** — проверка deliverable по TECH-xx criteria
 - **`$handoff`** — формирование handoff-конверта

package/domains/product/agents/ux_designer.md

@@ -43,6 +43,7 @@ early flow insight если Discovery указывает на complex flow.
 ## Используемые skills
 
 ### Обязательные (каждый раз)
+- **$karpathy-guidelines** — сначала думай, делай только нужное, правь точечно, работай от результата
 - **`$user-flow`** — user flow docs (шаги, decision points, exits, error states)
 - **`$design-brief`** — бриф для Designer / external visual designer
 - **`$gates`** — проверка deliverable по UX-01 criteria

package/locales/en/.agents/skills/karpathy-guidelines/SKILL.md

@@ -0,0 +1,325 @@
+---
+name: karpathy-guidelines
+description: Behavioral guidelines to reduce common LLM mistakes across all tasks — code, design, analytics, product. Apply always.
+license: MIT
+source: https://x.com/karpathy/status/2015883857489522876
+---
+
+# Karpathy Guidelines
+
+Behavioral guidelines to reduce common LLM mistakes, derived from [Andrej Karpathy's observations](https://x.com/karpathy/status/2015883857489522876).
+
+**Tradeoff:** These guidelines bias toward caution over speed. For trivial tasks, use judgment.
+
+---
+
+## 1. Think Before Coding
+
+**Don't assume. Don't hide confusion. Surface tradeoffs.**
+
+Before implementing:
+- State your assumptions explicitly. If uncertain, ask.
+- If multiple interpretations exist, present them — don't pick silently.
+- If a simpler approach exists, say so. Push back when warranted.
+- If something is unclear, stop. Name what's confusing. Ask.
+
+### Example: Hidden Assumptions
+
+**Request:** "Add a feature to export user data"
+
+❌ **What LLMs do** — silently assume format, fields, scope:
+
+```python
+def export_users(format='json'):
+    users = User.query.all()
+    with open('users.json', 'w') as f:
+        json.dump([u.to_dict() for u in users], f)
+    return f"Exported {len(users)} users"
+```
+
+✅ **What should happen** — surface assumptions and ask:
+
+```
+Before implementing, I need to clarify:
+
+1. Scope: Export all users or filtered subset? (privacy implications)
+2. Format: "export" — download file, background job, or API endpoint?
+3. Fields: Which user fields? (some might be sensitive)
+4. Volume: How many users? (affects approach)
+
+What's your preference?
+```
+
+### Example: Multiple Interpretations
+
+**Request:** "Make the search faster"
+
+❌ **What LLMs do** — silently pick one interpretation and write 200 lines of optimization.
+
+✅ **What should happen:**
+
+```
+"Make the search faster" could mean:
+
+1. Faster response time (< 100ms instead of 500ms)
+   → database indexes, cache | ~2h
+
+2. Handle more concurrent searches (throughput)
+   → async/await, connection pooling | ~4h
+
+3. Faster perceived speed (UX)
+   → show partial results immediately | ~3h
+
+Current search takes ~500ms. Which aspect matters most?
+```
+
+---
+
+## 2. Simplicity First
+
+**Minimum code that solves the problem. Nothing speculative.**
+
+- No features beyond what was asked.
+- No abstractions for single-use code.
+- No "flexibility" or "configurability" that wasn't requested.
+- No error handling for impossible scenarios.
+- If you write 200 lines and it could be 50, rewrite it.
+
+Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.
+
+### Example: Over-abstraction
+
+**Request:** "Add a function to calculate discount"
+
+❌ **What LLMs do** — Strategy pattern, ABC, dataclass, configurator:
+
+```python
+class DiscountStrategy(ABC):
+    @abstractmethod
+    def calculate(self, amount: float) -> float: pass
+
+class PercentageDiscount(DiscountStrategy):
+    def __init__(self, percentage: float):
+        self.percentage = percentage
+    def calculate(self, amount: float) -> float:
+        return amount * (self.percentage / 100)
+
+@dataclass
+class DiscountConfig:
+    strategy: DiscountStrategy
+    min_purchase: float = 0.0
+    max_discount: float = float('inf')
+
+# 40 more lines to use one formula
+```
+
+✅ **What should happen:**
+
+```python
+def calculate_discount(amount: float, percent: float) -> float:
+    """Calculate discount amount. percent should be 0-100."""
+    return amount * (percent / 100)
+```
+
+When to add complexity: only when multiple discount types are actually needed.
+
+### Example: Speculative Features
+
+**Request:** "Save user preferences to database"
+
+❌ **What LLMs do** — cache, validator, merge flag, notifications:
+
+```python
+class PreferenceManager:
+    def __init__(self, db, cache=None, validator=None):
+        ...
+    def save(self, user_id, prefs, merge=True, validate=True, notify=False):
+        # 50 lines for what could be 3
+```
+
+✅ **What should happen:**
+
+```python
+def save_preferences(db, user_id: int, preferences: dict):
+    db.execute(
+        "UPDATE users SET preferences = ? WHERE id = ?",
+        (json.dumps(preferences), user_id)
+    )
+```
+
+Add caching, validation, merging later — only when actually needed.
+
+---
+
+## 3. Surgical Changes
+
+**Touch only what you must. Clean up only your own mess.**
+
+When editing existing code:
+- Don't "improve" adjacent code, comments, or formatting.
+- Don't refactor things that aren't broken.
+- Match existing style, even if you'd do it differently.
+- If you notice unrelated dead code, mention it — don't delete it.
+
+When your changes create orphans:
+- Remove imports/variables/functions that YOUR changes made unused.
+- Don't remove pre-existing dead code unless asked.
+
+**The test:** Every changed line should trace directly to the user's request.
+
+### Example: Drive-by Refactoring
+
+**Request:** "Fix the bug where empty emails crash the validator"
+
+❌ **What LLMs do** — improve email validation beyond the bug fix, add username validation, change comment style, add docstring.
+
+✅ **What should happen — only the lines that fix empty email:**
+
+```diff
+  def validate_user(user_data):
+      # Check email format
+-     if not user_data.get('email'):
++     email = user_data.get('email', '')
++     if not email or not email.strip():
+          raise ValueError("Email required")
+
+      # Basic email validation
+-     if '@' not in user_data['email']:
++     if '@' not in email:
+          raise ValueError("Invalid email")
+
+      # Check username  ← untouched, not in scope
+      if not user_data.get('username'):
+          raise ValueError("Username required")
+
+      return True
+```
+
+### Example: Style Drift
+
+**Request:** "Add logging to the upload function"
+
+❌ **What LLMs do** — add type hints, change single quotes to double quotes, add docstring, reformat whitespace.
+
+✅ **What should happen — match existing style:**
+
+```diff
++ import logging
++
++ logger = logging.getLogger(__name__)
++
+  def upload_file(file_path, destination):
++     logger.info(f'Starting upload: {file_path}')
+      try:
+          with open(file_path, 'rb') as f:
+              data = f.read()
+
+          response = requests.post(destination, files={'file': data})
+
+          if response.status_code == 200:
++             logger.info(f'Upload successful: {file_path}')
+              return True
+          else:
++             logger.error(f'Upload failed: {file_path}, status={response.status_code}')
+              return False
+      except Exception as e:
+-         print(f"Error: {e}")
++         logger.exception(f'Upload error: {file_path}')
+          return False
+```
+
+Matched: single quotes, no type hints, existing boolean pattern, spacing style.
+
+---
+
+## 4. Goal-Driven Execution
+
+**Define success criteria. Loop until verified.**
+
+Transform tasks into verifiable goals:
+- "Add validation" → "Write tests for invalid inputs, then make them pass"
+- "Fix the bug" → "Write a test that reproduces it, then make it pass"
+- "Refactor X" → "Ensure tests pass before and after"
+
+For multi-step tasks, state a brief plan:
+
+```
+1. [Step] → verify: [check]
+2. [Step] → verify: [check]
+3. [Step] → verify: [check]
+```
+
+Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.
+
+### Example: Incremental with Verification
+
+**Request:** "Add rate limiting to the API"
+
+❌ **What LLMs do** — implement Redis, multiple strategies, config system, and monitoring in one 300-line commit with no clear verification steps.
+
+✅ **What should happen:**
+
+```
+Plan for rate limiting:
+
+1. Basic in-memory rate limiting (single endpoint)
+   Verify: 100 requests → first 10 succeed, rest get 429
+
+2. Extract to middleware (all endpoints)
+   Verify: existing endpoint tests still pass
+
+3. Redis backend (multi-server)
+   Verify: rate limit persists across app restarts
+
+4. Configuration (rates per endpoint)
+   Verify: /search allows 10/min, /users allows 100/min
+
+Start with step 1?
+```
+
+### Example: Reproduce Before Fixing
+
+**Request:** "The sorting breaks when there are duplicate scores"
+
+❌ **What LLMs do** — immediately change sort logic without reproducing the bug.
+
+✅ **What should happen:**
+
+```python
+# 1. First, write a test that reproduces the bug
+def test_sort_with_duplicate_scores():
+    scores = [
+        {'name': 'Alice', 'score': 100},
+        {'name': 'Bob', 'score': 100},
+        {'name': 'Charlie', 'score': 90},
+    ]
+    result = sort_scores(scores)
+    assert result[0]['score'] == 100
+    assert result[2]['score'] == 90
+    assert result[0]['name'] < result[1]['name']  # stable sort
+
+# Verify: test fails → bug reproduced
+
+# 2. Now fix with stable sort
+def sort_scores(scores):
+    return sorted(scores, key=lambda x: (-x['score'], x['name']))
+
+# Verify: test passes
+```
+
+---
+
+## Anti-Patterns Summary
+
+| Principle | Anti-Pattern | Fix |
+|---|---|---|
+| Think Before Coding | Silently assumes file format, fields, scope | List assumptions explicitly, ask for clarification |
+| Simplicity First | Strategy pattern for single discount calculation | One function until complexity is actually needed |
+| Surgical Changes | Reformats quotes, adds type hints while fixing bug | Only change lines that fix the reported issue |
+| Goal-Driven | "I'll review and improve the code" | "Write test for bug X → make it pass → verify no regressions" |
+
+## Key Insight
+
+The "overcomplicated" examples aren't obviously wrong — they follow design patterns and best practices. The problem is **timing**: they add complexity before it's needed.
+
+**Good code is code that solves today's problem simply, not tomorrow's problem prematurely.**

package/locales/en/.agents/skills/karpathy-guidelines/agents/claude.json

@@ -0,0 +1,22 @@
+{
+  "name": "karpathy-guidelines",
+  "display_name": "Karpathy Guidelines",
+  "description": "Behavioral guidelines to reduce common LLM mistakes — think before coding, simplicity first, surgical changes, goal-driven execution.",
+  "default_prompt": "Use $karpathy-guidelines when the task requires quality discipline: writing code, reviewing, refactoring, analysis, or design.",
+  "triggers": [
+    "karpathy-guidelines",
+    "karpathy",
+    "Karpathy Guidelines",
+    "think before coding",
+    "simplicity first",
+    "surgical changes",
+    "goal-driven execution"
+  ],
+  "capabilities": [
+    "quality",
+    "guidelines",
+    "best-practices"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/karpathy-guidelines/agents/copilot.json

@@ -0,0 +1,22 @@
+{
+  "name": "karpathy-guidelines",
+  "display_name": "Karpathy Guidelines",
+  "description": "Behavioral guidelines to reduce common LLM mistakes — think before coding, simplicity first, surgical changes, goal-driven execution.",
+  "default_prompt": "Use $karpathy-guidelines when the task requires quality discipline: writing code, reviewing, refactoring, analysis, or design.",
+  "triggers": [
+    "karpathy-guidelines",
+    "karpathy",
+    "Karpathy Guidelines",
+    "think before coding",
+    "simplicity first",
+    "surgical changes",
+    "goal-driven execution"
+  ],
+  "capabilities": [
+    "quality",
+    "guidelines",
+    "best-practices"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/karpathy-guidelines/agents/gemini.json

@@ -0,0 +1,22 @@
+{
+  "name": "karpathy-guidelines",
+  "display_name": "Karpathy Guidelines",
+  "description": "Behavioral guidelines to reduce common LLM mistakes — think before coding, simplicity first, surgical changes, goal-driven execution.",
+  "default_prompt": "Use $karpathy-guidelines when the task requires quality discipline: writing code, reviewing, refactoring, analysis, or design.",
+  "triggers": [
+    "karpathy-guidelines",
+    "karpathy",
+    "Karpathy Guidelines",
+    "think before coding",
+    "simplicity first",
+    "surgical changes",
+    "goal-driven execution"
+  ],
+  "capabilities": [
+    "quality",
+    "guidelines",
+    "best-practices"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/karpathy-guidelines/agents/openai.yaml

@@ -0,0 +1,8 @@
+interface:
+  display_name: "Karpathy Guidelines"
+  short_description: "Behavioral guidelines to reduce common LLM mistakes — think before coding, simplicity first, surgical changes, goal-driven execution."
+  default_prompt: "Use $karpathy-guidelines when the task requires quality discipline: writing code, reviewing, refactoring, analysis, or design."
+dependencies:
+  tools: []
+policy:
+  allow_implicit_invocation: true

package/locales/en/.agents/skills/karpathy-guidelines/agents/qwen.json

@@ -0,0 +1,22 @@
+{
+  "name": "karpathy-guidelines",
+  "display_name": "Karpathy Guidelines",
+  "description": "Behavioral guidelines to reduce common LLM mistakes — think before coding, simplicity first, surgical changes, goal-driven execution.",
+  "default_prompt": "Use $karpathy-guidelines when the task requires quality discipline: writing code, reviewing, refactoring, analysis, or design.",
+  "triggers": [
+    "karpathy-guidelines",
+    "karpathy",
+    "Karpathy Guidelines",
+    "think before coding",
+    "simplicity first",
+    "surgical changes",
+    "goal-driven execution"
+  ],
+  "capabilities": [
+    "quality",
+    "guidelines",
+    "best-practices"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/karpathy-guidelines/agents/skill.yaml

@@ -0,0 +1,25 @@
+version: 1
+name: "karpathy-guidelines"
+display_name: "Karpathy Guidelines"
+description: "Behavioral guidelines to reduce common LLM mistakes — think before coding, simplicity first, surgical changes, goal-driven execution."
+default_prompt: "Use $karpathy-guidelines when the task requires quality discipline: writing code, reviewing, refactoring, analysis, or design."
+triggers:
+  - "karpathy-guidelines"
+  - "karpathy"
+  - "Karpathy Guidelines"
+  - "think before coding"
+  - "simplicity first"
+  - "surgical changes"
+  - "goal-driven execution"
+capabilities:
+  - "quality"
+  - "guidelines"
+  - "best-practices"
+tools: []
+invocation:
+  explicit: true
+  implicit: true
+localization:
+  default_locale: "en"
+  available_locales:
+    - "en"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "code-ai-installer",
-  "version": "2.3.0",
+  "version": "2.4.0",
   "description": "Production-ready CLI to install code-ai agents and skills for multiple AI coding assistants.",
   "license": "MIT",
   "author": "Denis Harchenko",