Run all 3 task graders with strict in-range scores

Route inference API call through injected LiteLLM proxy

Emit exact [START]/[STEP]/[END] stdout tokens for validator

Add Hugging Face Spaces YAML config to root README

Align inference output to START/STEP/END and checklist env vars

Emit evaluator-friendly structured output in inference.py

Add multi-mode deployment metadata and server entrypoint

Add root pyproject.toml for openenv validate

Fix /reset to accept empty POST body

Tighten root submission packaging

Add root-level submission entrypoints

email-triage-env: hackathon hardening — scenarios, rubric, tests, API, packaging

Add EmailTriageEnv benchmark environment