claudify-agent 1.0.1 → 1.0.2

package/.claude/rules/mcp-auto-use.md

@@ -1,46 +1,46 @@
-# MCP Server Auto-Use
-
-## Automatic Triggers
-
-### user-context7 (Library/Framework Questions)
-**Trigger Keywords**: "how to", "implement", "best practice", "pattern", "library", "framework", "package"
-
-**Auto-Fetch Examples**:
-- User: "Add authentication" → Fetch: "Next.js auth patterns 2026"
-- User: "Create API endpoint" → Fetch: "REST API best practices"
-- User: "Use WebSocket" → Fetch: "WebSocket implementation guide"
-
-**Rule**: **ALWAYS** fetch docs BEFORE implementing unfamiliar patterns
-
-### Other MCP Servers
-
-| MCP Server | When to Use | Cost |
-|------------|-------------|------|
-| cursor-ide-browser | UI changes, visual verification | High |
-| user-github | PR/issue operations, git commands | Moderate |
-| user-ESLint | After EVERY code change (automatic) | Low |
-| user-chrome-devtools | Performance audits, Lighthouse | Moderate |
-| user-playwright | E2E testing, cross-browser | High |
-
-## Cost Hierarchy
-
-**Tier 1 (Use Liberally)**: user-ESLint, user-context7  
-**Tier 2 (Use When Needed)**: user-github, user-chrome-devtools  
-**Tier 3 (Use Sparingly)**: cursor-ide-browser, user-playwright
-
-## Rules
-
-1. Always try cheaper tools first (follow escalation protocol)
-2. Run user-ESLint automatically after code changes
-3. Fetch docs with user-context7 BEFORE implementing
-4. Use cursor-ide-browser ONLY for visual verification
-5. Use user-github for all GitHub operations
-
-## Patterns & Examples
-See `.cursor/docs/MCP-INTEGRATION-GUIDE.md` for complete patterns, chaining examples, and cost optimization strategies.
-
-## Fallback
-- Browser unavailable → Use built-in Browser tools
-- GitHub unavailable → Use Shell with gh CLI
-- ESLint unavailable → Use Shell with npm run lint
-- Context7 unavailable → Use WebSearch
+# MCP Server Auto-Use
+
+## Automatic Triggers
+
+### context7 (Library/Framework Questions)
+**Trigger Keywords**: "how to", "implement", "best practice", "pattern", "library", "framework", "package"
+
+**Auto-Fetch Examples**:
+- User: "Add authentication" → Fetch: "Next.js auth patterns 2026"
+- User: "Create API endpoint" → Fetch: "REST API best practices"
+- User: "Use WebSocket" → Fetch: "WebSocket implementation guide"
+
+**Rule**: **ALWAYS** fetch docs BEFORE implementing unfamiliar patterns
+
+### Other MCP Servers
+
+| MCP Server | When to Use | Cost |
+|------------|-------------|------|
+| claude-in-chrome | UI changes, visual verification | High |
+| github | PR/issue operations, git commands | Moderate |
+| eslint | After EVERY code change (automatic) | Low |
+| chrome-devtools | Performance audits, Lighthouse | Moderate |
+| playwright | E2E testing, cross-browser | High |
+
+## Cost Hierarchy
+
+**Tier 1 (Use Liberally)**: eslint, context7  
+**Tier 2 (Use When Needed)**: github, chrome-devtools  
+**Tier 3 (Use Sparingly)**: claude-in-chrome, playwright
+
+## Rules
+
+1. Always try cheaper tools first (follow escalation protocol)
+2. Run eslint automatically after code changes
+3. Fetch docs with context7 BEFORE implementing
+4. Use claude-in-chrome ONLY for visual verification
+5. Use github for all GitHub operations
+
+## Patterns & Examples
+See `.claude/docs/MCP-INTEGRATION-GUIDE.md` for complete patterns, chaining examples, and cost optimization strategies.
+
+## Fallback
+- Browser unavailable → Use playwright MCP, or WebFetch for static pages
+- GitHub unavailable → Use Bash with gh CLI
+- ESLint unavailable → Use Bash with npm run lint
+- Context7 unavailable → Use WebSearch

package/.claude/rules/parallel-exploration-auto.md

@@ -1,31 +1,30 @@
-# Parallel Exploration Auto
-
-## Automatic Triggers
-
-**Launch 2-4 parallel subagents when**:
-- Task spans >3 areas/directories
-- Complexity >10 steps or >5 files
-- Bug investigation (unknown root cause)
-- Agent unfamiliarity detected (>3 questions)
-
-## Configuration
-
-```yaml
-Task:
-  subagent_type: "explore"
-  model: "fast"  # ALWAYS for exploration
-  readonly: true
-  run_in_background: true
-  timeout: 120s
-```
-
-## Rules
-
-1. Default to 2-3 subagents, scale to 4 for complex tasks
-2. ALWAYS use `model="fast"` (90% of cases)
-3. Consolidate findings before planning
-4. See `.cursor/skills/parallel-exploration/SKILL.md` for patterns
-
-## Anti-Patterns
-
-❌ >4 subagents, ❌ capable model for exploration, ❌ no consolidation
+# Parallel Exploration Auto
+
+## Automatic Triggers
+
+**Launch 2-4 parallel subagents when**:
+- Task spans >3 areas/directories
+- Complexity >10 steps or >5 files
+- Bug investigation (unknown root cause)
+- Agent unfamiliarity detected (>3 questions)
+
+## Configuration
+
+```yaml
+Agent:
+  subagent_type: "Explore"
+  model: "haiku"  # ALWAYS for exploration
+  run_in_background: true
+  timeout: 120s
+```
+
+## Rules
+
+1. Default to 2-3 subagents, scale to 4 for complex tasks
+2. ALWAYS use `model="haiku"` (90% of cases)
+3. Consolidate findings before planning
+4. See `.claude/skills/parallel-exploration/SKILL.md` for patterns
+
+## Anti-Patterns
+
+❌ >4 subagents, ❌ capable model for exploration, ❌ no consolidation