atul120411/sentinel-os-agentic-benchmark

tighten inference contract and docs

Keep grader scores inside open interval

Align task ids and baseline with phase 2 contract

Remove legacy tasks package after standardization

Standardize task registry to top-level tasks module

Add task metadata and grader alias compatibility

Align layout with canonical OpenEnv server structure

Add validator-friendly task registry exports

Add validator grader compatibility surface

Return tasks as plain list for validator

Use ECR mirror for Python base image

Harden Docker build for validator

Expose task grader discovery endpoints

Tighten OpenEnv spec compliance

final: sentinel-os v2.2 production-grade benchmark

final: structured graders + phase2 ready

final: structured graders + phase2 ready

add: uv lockfile for reproducibility

fix: openenv compliance (main entrypoint)

fix: align runtime with POST /reset

fix: add POST /reset support

remove redundant nested repo and update README

update README for HF deployment

fresh clean submission