feat(sn60): evidence-first DeepSeek depth miner (jimcody1995-20260718-01)

[jimcody1995]

Summary

• Challenger for current king bohdansolovie / PR feat(sn60): DeepSeek depth-audit miner (bohdansolovie-20260717-01) #156 (deepseek-ai/DeepSeek-V3.2-TEE, 6 TP / ~9% detection).
• Pipeline: static risk-rank → map triage → single-file deep audits (value / auth / oracle lenses) → verify pass.
• No structural-probe findings (king emitted many FPs at ~3.5% precision). All findings are LLM-justified.
• Languages: Solidity, Vyper, Cairo, Rust. Phala TEE timings + Chutes-sealed key.

Why this should beat the field

• Deeper whole-contract context per call vs multi-file dumps.
• Verify pass filters unsupported candidates before scoring.
• Caps emitted findings to protect precision while maximizing true positives.

Bundle

agent.py · agent_manifest.json · submission.json · sealed_inference_key

Test plan

• Intake screening → kata:pending
• Round scores vs king + open challengers
• Strictly beats on projects passed / true positives

[carlos4s]

I closed this PR because it failed the round-start execution screener.

Your PR passed intake earlier, but before scoring I run the candidate once on a real benchmark project to make sure it can execute and return valid report JSON.

Screener project: code4rena_loopfi_2025_02

Reason:

• SN60 screening execution did not complete successfully: sealed-room run rejected: room run failed: Remote end closed connection without response
• SN60 screening report must be a JSON object.

This candidate did not enter the main round scoring, so this is marked kata:invalid rather than kata:losing.