Purpose: This file provides Claude with expert-level guidance for generating comprehensive, high-quality test suites using CoreStory’s MCP server. When a user asks Claude to generate tests, Claude should reference these principles and workflows.

Core Principles

1. Always Use CoreStory Context First

Before generating any tests:

• Use CoreStory tools to understand the actual codebase structure
• Analyze existing code patterns and conventions
• Identify dependencies and integration points
• Review business logic and data flows
• Don’t assume or hallucinate code structure

Workflow:

1. list_projects → confirm project_id with user
2. get_project → verify project details
3. create_conversation → establish context for this test generation session
4. send_message → ask CoreStory about the specific module/feature to test
5. Generate tests based on ACTUAL specifications or source code, not assumptions

2. The T.E.S.T. Framework

Every test generation request should extract:

• Target: What specific code needs testing?
• Expectations: What test types and coverage goals?
• Specifications: What are the business rules and acceptance criteria?
• Technical: What frameworks, patterns, and constraints?

If any component is unclear, ASK before generating.

3. Progressive Refinement Over Bulk Generation

Don’t: Generate 100 tests in one shot Do: Generate in rounds with validation:

Round 1: Happy path tests (10-15 tests)
   ↓ User reviews
Round 2: Edge cases based on feedback (15-20 tests)
   ↓ User reviews
Round 3: Error scenarios and integration tests (10-15 tests)
   ↓ User reviews
Round 4: Performance/security tests if needed (5-10 tests)

This approach:

• Allows course correction
• Builds user confidence
• Prevents wasted effort
• Ensures alignment with expectations

CoreStory Integration Workflow

Step 1: Project Discovery

1. If user mentions project_id → use it
2. If not → Call list_projects
3. Present options to user: "I see you have these projects: [list]. Which one should we work on?"
4. Call get_project to confirm and show project details
5. Call get_project_stats to check if codebase is fully indexed

Step 2: Code Understanding

1. Create conversation: create_conversation(project_id, title="Test Generation for {feature}")
2. Query CoreStory about the target:
   - "What does the {module} do?"
   - "What are the dependencies of {component}?"
   - "Show me the business logic in {file}"
   - "What edge cases should I consider for {feature}?"
3. Use send_message to get contextual information
4. Build mental model of the code before generating tests