Skip to content

PRE_SUBMISSION_CHECKLIST.md

Latest commit

 

History

History
148 lines (116 loc) · 4.12 KB

PRE_SUBMISSION_CHECKLIST.md

File metadata and controls

148 lines (116 loc) · 4.12 KB
Raw

Pre-Submission Checklist - PASSED ✅

Meta PyTorch Hackathon x Scaler School of Technology, Round 1

Project: Ultra Instinct (Sentinel Environment)
Date: 2026-04-10
Validation Script: validate_submission.py
Result: ✅ 57/57 CHECKS PASSED

✅ Pre-Submission Checklist

1. HF Space Deploys ✅

2. Automated Ping to Space URL ✅

  • GET /health returns 200
  • Responds to reset() endpoint
  • All endpoints responsive

3. OpenEnv Spec Compliance ✅

  • openenv.yaml valid with name, sdk, metadata
  • SDK is "docker"
  • Typed Pydantic models (models.py)
  • step() endpoint works
  • reset() endpoint works
  • state() endpoint works
  • 3 tasks defined in metadata

4. Dockerfile Builds ✅

  • FROM instruction present
  • WORKDIR set
  • COPY instructions present
  • EXPOSE/CMD present
  • Uses Python/PyTorch base image

5. Baseline Reproduces ✅

  • inference.py runs without syntax errors
  • Produces structured output ([START], [STEP], [END])
  • All environment variables have defaults
  • Exit code 0 guaranteed

6. 3+ Tasks with Graders ✅

  • basic-injection (easy) - score=0.60
  • social-engineering (medium) - score=0.60
  • stealth-exfiltration (hard) - score=0.29
  • All scores in [0.0, 1.0] range
  • All tasks graded successfully

7. Mandatory Environment Variables ✅

  • API_BASE_URL defined with default
  • MODEL_NAME defined with default
  • HF_TOKEN defined

8. Inference Script Requirements ✅

  • Named inference.py in root directory
  • Uses OpenAI Client (from openai import OpenAI)
  • Uses client.chat.completions.create for LLM calls
  • Emits [START] with task, env, model fields
  • Emits [STEP] with step, action, reward fields
  • Emits [END] with success, steps, score fields

9. Infra Restrictions ✅

  • Runtime < 20min (validated with MAX_STEPS=2)
  • Compatible with vCPU=2, memory=8GB
  • No excessive resource requirements

📊 Validation Results

======================================================================
  VALIDATION SUMMARY
======================================================================
  Passed: 57
  Failed: 0
======================================================================

  ✅ ALL CHECKS PASSED - READY FOR SUBMISSION

Categories Validated:

  1. File Structure (5/5) ✅
  2. Environment Variables (5/5) ✅
  3. OpenAI Client Usage (3/3) ✅
  4. Structured Logging Format (6/6) ✅
  5. OpenEnv Spec Compliance (6/6) ✅
  6. HF Space Health (3/3) ✅
  7. Endpoint Compliance (13/13) ✅
  8. Three Tasks with Graders (4/4) ✅
  9. Inference Script Validation (5/5) ✅
  10. Dockerfile Validation (5/5) ✅

🚀 Deployment Status

Component Status Details
GitHub ✅ Pushed Commit: 9dc2d1a
Hugging Face ✅ Deployed 166 files uploaded
Space Health ✅ Healthy v1.1.0
Validation ✅ Passed 57/57 checks
Tasks ✅ 3/3 All graded

📝 How to Use

Before ANY Commit or Deployment:

uv run python validate_submission.py

MUST see: ✅ ALL CHECKS PASSED - READY FOR SUBMISSION

If ANY check fails, the script will:

  • Show which checks failed in RED
  • Exit with code 1
  • Prevent deployment

Deployment:

uv run python deploy-hf-now.py

Submit to Hackathon:

  1. GitHub Repo: https://github.com/Pranava-Kumar/Sentinel-Env
  2. HF Space: https://huggingface.co/spaces/PranavaKumar09/sentinel-env

🔒 IP Protection

  • ✅ jailbreak-prompts/ NOT in git (gitignored)
  • ✅ 114 real attack prompts only in HF Space container
  • ✅ Secret sauce protected - not in public repository
  • ✅ Only deployment wrapper in git, actual prompts in HF Space only

Status: ✅ READY FOR SUBMISSION
Last Validated: 2026-04-10
Next Validation: Run validate_submission.py before any changes