The fourth Hunter engine — created inside the DataHub hackathon window, DataHub-native from this first commit. It hunts hackathons and prize challenges: scouts sweep the official platforms (Devpost, MLH), and one deterministic, fail-closed gate decides what may be trusted:
- official_listing_confirmed — the candidate's own source resolves to the real platform, never a lookalike "prize portal" or a roundup article
- deadline_is_future — a parseable submission deadline that hasn't passed
- prize_documented — a real, documented prize pool ("exposure" doesn't gate)
- eligibility_parseable — the rules say who may enter
plus the substrate's inherited scam-list / allowlist / domain-age / multi-source / scam-report machinery. Debate stances (Advocate, Skeptic, Strategist arguing theme-fit vs time budget) are transcript, never verdict.
Built on the shared hunter-engine substrate — the same scout/gate/debate/mission machinery behind Crypto, Collectible, and Free Money Hunter AI, arbitrated by Opportunity [Agency AI], operated from Entropy OS.
DataHub-native: every gate verdict this engine records is emitted to a real
DataHub instance as governed datasets with assertion verdicts, via Veritas's
orgs/datahub_opportunity_emit.py — no adapter code, the emit path
generalizes across the family.
python3 -m venv .venv && .venv/bin/pip install -r requirements.txt
.venv/bin/pytest -q # the gate, proven offline
.venv/bin/python cli.py day --no-voice # one full cycle: sweep → gate → digest → mission
.venv/bin/python cli.py serve # dashboard at http://127.0.0.1:8016A MoreSalamander StudioLabs production — the model proposes, the gate disposes.