Feature Request: Project-aware initialization and workflow learning

Problem Statement

Currently, skills, MCP tools, and sub-agents are powerful features for structuring development workflows, but they lack a systematic way to:

• Initialize project-specific configurations
• Learn from user interaction patterns
• Automatically suggest optimizations based on usage

Proposed Solution

Phase 1: Intelligent Initialization

When starting work on a new project, Claude Code could:

1. Project Profiling (via questionnaire or conversational flow)

• Detect tech stack (frameworks, languages, build tools)
• Understand team size and development methodology
• Identify common workflows (testing, deployment, code review)

1. Smart Recommendations

• Suggest relevant skills based on project characteristics
• Recommend MCP tools that match the tech stack
• Propose sub-agent configurations for common tasks
• Generate .claude/ configuration templates

1. Example Flow

```
Claude: I detected this is a Laravel project with Vue frontend.
Recommended setup:

• Skills: /commit-flow, /deploy-staging, /run-tests
• MCP: context7 (for Laravel docs), exa (for package research)
• Sub-agents: test-runner, code-reviewer

Would you like me to initialize these?
```

Phase 2: Continuous Learning & Iteration

1. Pattern Recognition

• Track high-frequency operation sequences
• Identify repetitive multi-step workflows
• Detect common file access patterns

1. Proactive Suggestions

• "I noticed you always run tests before committing. Create a /pre-commit skill?"
• "You frequently grep then read files in the same module. Should I create an explore agent for this?"
• "This 5-step deployment sequence appears weekly. Save as /deploy skill?"

1. Habit Recording

• After completing a workflow, ask: "Save this as a reusable command/skill?"
• Learn preferred tool choices (e.g., user prefers grep over glob)
• Adapt verbosity and explanation depth based on user feedback

1. Configuration Evolution

• Automatically update .claude/workflows.md with learned patterns
• Version control for workflow configurations
• Share learned patterns across team (optional)

Benefits

1. Lower Barrier to Entry: New users get project-appropriate setup automatically
2. Efficiency Gains: Reduce repetitive manual configuration
3. Knowledge Capture: Workflows become documented and shareable
4. Continuous Improvement: System gets better with usage
5. Team Alignment: Standardize workflows across team members

Implementation Ideas

• Store learned patterns in .claude/learned-patterns.json
• Use LLM to analyze conversation history for pattern extraction
• Provide /review-workflows command to see suggestions
• Add /export-config to share configurations
• Implement privacy controls for what gets learned/shared

Use Case Example

Week 1: User works on Laravel project, manually runs pint → test → commit each time

Week 2: Claude suggests: "Create /commit skill for this sequence?"

Week 3: User accepts, now uses /commit shortcut

Week 4: Claude notices user always checks queue status after deployment, suggests adding to /deploy skill

Result: Personalized, evolving workflow that matches actual usage patterns

Related Context

This suggestion comes from real-world usage where the power of skills/MCP/sub-agents is evident, but the configuration and discovery process could be more intuitive and adaptive.

---

Submitted by: User feedback via Claude Code session
Project Context: Laravel-based ticket booking system with modular architecture