debug: Phase 1 MCP diagnostics - cache clearing and environment snapshot

[lgallard]

Phase 1: Quick Elimination Tests for MCP Server Connectivity

This PR implements the first phase of systematic debugging to fix the Terraform MCP server connectivity issue.

Problem

• Terraform MCP server: "status": "failed" ❌
• Context7 MCP server: "status": "connected" ✅
• ECR repository works with identical configuration
• Same PAT token used in both repositories

Phase 1 Diagnostics Added

🧹 Cache Clearing

• Complete NPM cache cleaning (npm cache clean --force)
• NPX cache directory removal (~/.npm/_npx)
• Cache log cleanup (~/.npm/_cacache, ~/.npm/_logs)

📊 Environment Snapshot

• Node.js, NPM, NPX versions
• Runner OS and system info
• Disk space and memory status
• Network connectivity test (npmjs.org)

🔍 MCP Package Testing

• Terraform MCP server package accessibility
• Context7 MCP server package accessibility
• Pre-installation NPX verification

Expected Outcomes

1. If cache clearing fixes it: MCP server status becomes "connected"
2. If environment differs: Diagnostic output shows differences vs ECR repo
3. If packages inaccessible: Network/registry issues identified
4. If still failing: Proceed to Phase 2 (Canary Swap test)

Next Steps

• Test workflow on this branch
• Compare diagnostic output with ECR repository
• If Phase 1 doesn't solve it, proceed to Phase 2 (content vs environment isolation)

This systematic approach will either solve the issue quickly or provide definitive data for Phase 2.

[lgallard]

Closing this diagnostic PR as it's superseded by the successful fix in PR #232. The cache clearing and diagnostics were helpful for debugging, but the root cause was identified as tool permission mismatches with the HashiCorp Terraform MCP server.