Team NIRVIKALP (HackIndia Spark-11): ANTARYA — Dukan Ka Dimaag (80% eval pass rate, 4 ADL cycles, validated agentspec v0.3.0)

[Arjundas08]

🏆 HackIndia Spark-11 Submission (CBIT Hyderabad)

Team NIRVIKALP — ANTARYA (अंतर्या) – Dukan Ka Dimaag

Autonomous Kirana Store Operating System powered by Mutagent ADL & 3-Brain Tri-Engine.

📁 What's in this PR (submissions/team-nirvikalp/)

• ✅ agentspec.yaml — validated against agentspec.mutagent.io/v0.3.0 ([validate-spec] PASS)
• ✅ eval_suite.json — 20 Kirana store test scenarios across 6 business categories
• ✅ 4 Real HTTP Evaluation Cycles: 45% → 50% → 75% → 80% (16/20 Live Gemini Pass Rate)
• ✅ Traces & Transcripts: 80 trace files (20 cases × 4 cycles) + 3 full session JSONL transcripts
• ✅ Product Feedback: 3 actionable CLI feedback items filed via mutagent feedback send
• ✅ Custom Extension: *retail-health evaluator skill for Kirana store risk auditing (extensions/retail-health/SKILL.md)
• ✅ Architecture & Architecture Diagram: architecture.md + antarya_hurdle_ml_architecture.png
• ✅ Autonomous Store Manager: 7-stage goal-directed loop (server/services/storeManager.js)

🔗 Primary Repository: https://github.com/Arjundas08/antarya3