AGENT SKILL

game-theoretic-swarm-deliberator

Orchestrate multi-persona LLM swarms to deliberate multi-domain strategic actions and resolve game-theoretic payoff matrices using Borda, weighted majority, and consensus voting. Do not use for linear single-agent workflows.

License 许可
MIT
最近更新
2026-11-15
Stars
★ 0
Forks
0

01 前言 Front Matter

name
game-theoretic-swarm-deliberator
description
Orchestrate multi-persona LLM swarms to deliberate multi-domain strategic actions and resolve game-theoretic payoff matrices using Borda, weighted majority, and consensus voting. Do not use for linear single-agent workflows.

02 资源文件 2

03 SKILL.md 正文

Game-Theoretic Swarm Deliberator: Multi-Persona Action & Payoff Voting Engine

game-theoretic-swarm-deliberator is the multi-agent consensus and game-theoretic voting engine distilled from the Strategify simulation harness. It orchestrates parallel domain personas (e.g. Defense, Epidemiology, Finance, Diplomacy, Legal) to deliberate strategic options and evaluate candidate payoff matrices using formal social choice voting rules.

Every deliberation session follows a deterministic 5-stage progression:

[1. Snapshot & Persona Fan-Out] → [2. Parallel Deliberation] → [3. Payoff Matrix Formulation] → [4. Voting & Consensus Synthesis] → [5. Verification & Action Dispatch]

See CARD.md for the companion summary card, voting rule cheat sheet, and invariant checklist.


The 5-Stage Deliberation Progression

Stage 1: Snapshot & Persona Fan-Out

Ingest the multi-domain simulation state snapshot and configure persona priority weights:

  1. Domain Minister Personas:
    • Defense (Priority Weight: 1.2x) — Focuses on readiness, spectrum allocation, EW jamming, and deterrent force posture.
    • Epidemiology (Priority Weight: 1.1x) — Focuses on transmission rates ($R_0$), NPI lockdowns, and vaccine R&D.
    • Finance (Priority Weight: 1.0x) — Focuses on GDP growth, trade tariffs, supply chain chokepoint resilience, and inflation.
    • Diplomacy (Priority Weight: 1.0x) — Focuses on bilateral tensions, treaty compliance, multilateral summits, and alliance pacts.
  2. Provider Routing: Query live providers (Ollama, OpenAI, Anthropic) or engage deterministic rule-based heuristic fallbacks.

Completion criterion: Persona roster initialized with domain priority weights and state snapshot summary.


Stage 2: Parallel Deliberation & Proposal Generation

Each persona independently reviews the domain snapshot and emits a structured proposal:

  • persona_name: Name and role identifier.
  • domain: Functional portfolio.
  • recommended_action: Proposed tactical or policy action.
  • confidence_score: Float between $0.0$ and $1.0$.
  • reasoning_chain: Step-by-step rationale citing snapshot metrics.

Completion criterion: Structured AgentProposal objects harvested from 100% of participating personas.


Stage 3: Payoff Matrix & Action Candidate Formulation

Formulate the normal-form game or strategic candidate vector:

  • Construct candidate action profiles $A = (a_1, a_2, \dots, a_k)$ or payoff candidate matrices.
  • Align candidate choices against current adversary threat models and environmental stress factors.

Completion criterion: Candidate action profile set or payoff candidate matrix array populated.


Stage 4: Voting & Consensus Synthesis

Evaluate candidate options using the designated social choice aggregation rule:

  1. Borda Count Voting (voting_rule="borda"):
    • Each persona ranks candidates from best to worst.
    • Scores are assigned: $(N - ext{rank}) imes ext{priority_weight}$.
    • The candidate with the maximum cumulative Borda score is selected as the winning action.
  2. Weighted Majority Voting (voting_rule="majority"):
    • Each persona casts its top preference weighted by its priority multiplier.
    • The candidate receiving the plurality of weighted votes wins.
  3. Unanimous Consensus (voting_rule="unanimous"):
    • Requires top-choice unanimity across all personas; defaults to safest defensive candidate upon deadlock.

Completion criterion: Winning strategic payload synthesized and composite consensus score calculated ($\in [0, 1]$).


Stage 5: Verification & Action Dispatch

  1. Validate consensus action vector against structural safety invariants and treaty compliance checks.
  2. Dispatch action vector to simulation stepping engine or external execution environment.
  3. Record deliberation transcript and consensus score to execution session history.

Completion criterion: Deliberation result validated and dispatched with zero unhandled exceptions.


Voting Rule Matrix Reference

Voting Rule Mathematical Formulation Best Used For Deadlock Behavior
Borda Count $ ext{Score}(c) = \sum_{p} (N - ext{rank}_p(c)) \cdot w_p$ Multi-criteria trade-offs with conflicting priorities Max score tie-breaker
Weighted Majority $ ext{Votes}(c) = \sum_{p: ext{top}(p)=c} w_p$ High-tempo tactical decision-making Highest weight persona
Unanimous $ orall p_i, p_j: ext{top}(p_i) = ext{top}(p_j)$ High-stakes nuclear/escalation thresholds Conservative fallback

The Three Foundational Pillars

1. The Visual Brief

Synthesize findings, structural matrices, and execution DAGs into a self-contained HTML visual brief written to %TEMP%\game-theoretic-swarm-deliberator-<timestamp>.html (or /tmp/game-theoretic-swarm-deliberator-<timestamp>.html on Unix) loading Tailwind CSS and Mermaid.js via CDN. Present the absolute clickable path to the user for visual review.

2. The Mandatory Checkpoint

When proposing structural modifications, refactors, or execution pipelines, always author an implementation_plan.md artifact with RequestFeedback: true. The agent must STOP and wait for explicit human confirmation before executing mutating actions.

3. Explicit Anti-Patterns

Rigid behavioral boundaries must be maintained to prevent catastrophic failure modes.


Anti-Patterns

  • Monolithic Persona Blending — Collapsing distinct domain perspectives into a single generic prompt, eroding domain tension and realistic trade-off modeling.
  • Unweighted Aggregation — Treating critical life-safety domains (Epidemiology, Defense) identically to soft diplomacy during existential crises.
  • Silent API Failure — Crashing the swarm when an external LLM endpoint times out instead of engaging deterministic rule-based heuristic fallbacks.
  • Unranked Payoff Guessing — Choosing payoff candidates without structured social choice aggregation.

04 同仓库更多 Skills

SpectreDeath
SpectreDeath/Brain-Harness

ui-behavior-primitive-architect

Architect, audit, and verify zero-dependency headless UI behavior primitives, WebMCP agent bridges, WAI-ARIA/SSR/RTL tri-modal checks, and framework expert routing. Do not use for static CSS theme styling or backend database migrations.

★ 0
SpectreDeath
SpectreDeath/Brain-Harness

trustgraph-ontology-governor

Govern knowledge graphs, SHACL-AF triple policies, OntologyRAG subclass validation, PROV-O dual named graphs, and Laya entropy decision scoring. Do not use for relational SQL queries or generic web scraping.

★ 0
SpectreDeath
SpectreDeath/Brain-Harness

shim-harness-governor

Govern universal CLI agent loops, 4-tier tool permissions, ancestor-symlink workspace sandboxes, deterministic-first command routing, collision-safe MCP namespacing, and layered skill loading. Do not use for static UI styling or external SQL analytics.

★ 0
SpectreDeath
SpectreDeath/Brain-Harness

graphiti-zep-memory-governor

Govern bi-temporal knowledge graphs with Graphiti future-edge clamping, MinHash/LSH entity dedup, RRF/MMR reranking, and Zep typed ontologies. Do not use for static SQL joins or generic web crawling.

★ 0
SpectreDeath
SpectreDeath/Brain-Harness

agora-mcp-workbench-architect

Architect, audit, and orchestrate Agora Workbench MCP servers, loopback sidecars, unified state graphs, gateway/dispatcher connectors, and policy-enforced data lakes. Do not use for static frontend CSS styling or non-MCP desktop GUI automation.

★ 0
SpectreDeath
SpectreDeath/Brain-Harness

skill-eval-assurance-runner

Evaluate and assure agent skills using small-N Wilson confidence intervals, unbiased pass^k reliability, deterministic trajectory graders, preflight boundary honesty, and sealed digest verification. Do not use for subjective prose review or unconstrained production deployment.

★ 0