evalgate v1.0.0

LLM prompt eval gate for CI.

• 8 assertion types (contains → JSON Schema → LLM judge)
• 4 providers over raw fetch: Anthropic, OpenAI, OpenRouter, mock (offline dry-runs)
• Sticky PR comment updated in place, job summary, gate with exit codes
• npx @arthurcarlson/evalgate run --dry-run to try it free

🤖 Generated with Claude Code