| name | hypothesis-ranked-tournament |
| description | Judge one ranked-tournament matchup between two top frontier hypotheses. |
hypothesis-ranked-tournament
Goal:
- Judge one ranked-tournament matchup between two top frontier hypotheses.
Inputs:
research_plan/RESEARCH_PLAN.json
- two
hypotheses/<id>/HYPOTHESIS.json artifacts with completed review bundles
Outputs:
- one
TournamentMatchContract result for a ranked tournament
Context Loading:
- Open
skills/shared-references/schema-index.md.
- Read
packages/agent_contracts/ranking.py and confirm the exact TournamentMatchContract shape before emitting the ranked result.
- Read
research_plan/RESEARCH_PLAN.json.
- Read both hypotheses and their review bundles.
- Confirm that the match strategy is
ranked_tournament.
Execution Prompt Contract:
- System Intent:
- You are running a deeper comparative judgment for two high-ranked hypotheses whose relative ordering matters.
- Required Reasoning Focus:
- Compare novelty, mechanism quality, evidential grounding, feasibility, and expected downstream value.
- When a debate-style internal comparison helps, keep it short and converge to one winner.
- Do Not Do:
- Do not leave the match undecided.
- Do not ignore review weaknesses that materially change the ordering.
- Output Shape:
- Emit one
TournamentMatchContract.
match_strategy: ranked_tournament
winner_id: one of the two hypothesis IDs
reasoning: concise but specific
debate_turns: optional short internal debate transcript when useful
Execution Steps:
- Open
skills/shared-references/schema-index.md, then read packages/agent_contracts/ranking.py before emitting the ranked match artifact.
- Compare the two frontier hypotheses for ranked ordering.
- Choose exactly one winner.
- Emit one valid ranked-tournament match result.
Completion Rule:
- This skill is complete only when one valid ranked match result has been produced with a non-empty winner and reasoning.