debate-hall-auto

name	debate-hall-auto
description	Automated Wind/Wall/Door debate orchestration with tier-based model selection.
triggers	["run debate","resolve question","automated debate","resume debate","tier selection"]
allowed-tools	["Read","mcp__debate-hall__run_debate","mcp__debate-hall__resolve_question","mcp__debate-hall__resume_debate","mcp__debate-hall__search_decisions"]

===DEBATE_HALL_AUTO===

META: TYPE::SKILL[FOCUSED] VERSION::"1.0" PURPOSE::"Automated debate orchestration via run_debate and resolve_question" ROUTER::debate-hall

§1::TOOLS

RUN_DEBATE::[ PURPOSE::"Single-call automated Wind/Wall/Door debate", PARAMS::[ topic::REQUIRED[question_or_decision], tier::"fast"|"standard"|"premium"[default:standard], thread_id::OPTIONAL[auto_generated_if_omitted], compression_tier::"none"|"basic"|"aggressive"|"ultra", primer_tier::"none"|"literacy"|"standard"|"advanced" ], RETURNS::synthesis+turn_count+status ]

RESOLVE_QUESTION::[ PURPOSE::"Highest-level API - debate + decision record in one call", PARAMS::[topic::REQUIRED,tier::OPTIONAL,thread_id::OPTIONAL], RETURNS::DecisionRecord[thread_id,synthesis,rationale,validation,provenance] ]

RESUME_DEBATE::[ PURPOSE::"Continue PAUSED debates after provider timeouts", PARAMS::[thread_id::REQUIRED,tier::OPTIONAL], USE_WHEN::status="PAUSED" ]

SEARCH_DECISIONS::[ PURPOSE::"Find past decisions by topic", PARAMS::[query::REQUIRED,limit::10,min_score::0.0], RETURNS::ranked_results[thread_id,topic,synthesis,score] ]

§2::TIER_CONFIGURATIONS

FAST::[ COST::~$0.02, MODELS::[gemini-3-flash,gpt-5.1-codex-mini,claude-haiku-4.5], MAX_TURNS::6, CONSENSUS::false, USE_FOR::[exploration,low_stakes,rapid_iteration] ]

STANDARD::[ COST::~$0.30, MODELS::[gemini-3-pro,gpt-5.2-codex,claude-sonnet-4.5], MAX_TURNS::12, REFINEMENT_LOOPS::4, USE_FOR::[design_decisions,medium_complexity] ]

PREMIUM::[ COST::~$0.60, MODELS::[claude-opus-4.5,gpt-5.2-pro,gemini-3-pro], MAX_TURNS::16, REFINEMENT_LOOPS::5, CONSENSUS::true, USE_FOR::[production_decisions,high_stakes] ]

§2.1::AGENT_ROLES // Restored from monolith §5

TIER_1_BASIC::[ AGENTS::[wind-agent,wall-agent,door-agent], BEHAVIOR::[explores_obvious_paths,balanced_judgment,balanced_integration], MAPPED_TO::fast_tier ] TIER_2_SPECIALIST::[ PATHOS::[ideator[minimal_elegant],edge-optimizer[hidden_vectors]], ETHOS::[validator[cold_truth],critical-engineer[production_readiness]], LOGOS::[synthesizer[breakthrough_1+1=3],technical-architect[system_design]], MAPPED_TO::standard_and_premium_tiers ] MAPPING::specialists→cognition_role[PATHOS→Wind,ETHOS→Wall,LOGOS→Door]

§3::EXAMPLES

QUICK::run_debate("Redis vs Memcached for sessions?",tier:"fast") STANDARD::run_debate("API versioning strategy?",tier:"standard") PREMIUM::resolve_question("Migrate PostgreSQL to CockroachDB?",tier:"premium")

GRAVITY_HINTS::[ fast::[naming_conventions,ui_tweaks,tooling_choices], standard::[api_design,module_boundaries,testing_strategy], premium::[security_architecture,data_model,external_contracts] ]

§4::BEST_PRACTICES TIER_SELECTION::"Match tier to decision gravity" ESCALATION::"If fast insufficient, re-run with premium" THIRD_WAY::"Best debates synthesize what neither Wind nor Wall proposed alone"

===END===