⚡ v1.3.0  ·  Configurable  ·  Auditable  ·  Cost-Aware

Gang

Configurable Multi-Agent Business Committee for Claude Code.
One command. Expert analysis. Evidence-backed verdict.

/gang run
📖 Full Documentation & Deep Dive →

6-Stage Pipeline

1
INIT
Scan + Evidence Population
Codebase Scan evidence.json Quality Mode Score Rubric Web Research
2
THINK
Committee + Parallel Experts
7 Experts Evidence Protocol Assumptions Light / Deep
3
DEBATE
Cross-Review · 4 Modes
all-vs-all selective relevance focused
4
SCORE
Rubric-Anchored Synthesis
Rubric 1–10 Evidence IDs Weighted Avg Confidence
5
ADVISE
CEO/CTO · Guardrails
GO CONDITIONAL-GO NO-GO Opus Model
6
DELIVER
GO Package Output
BRD Architecture Charter Risk Register API Contracts
INIT
THINK
DEBATE
SCORE
ADVISE
DELIVER

The Committee

PM Lead
Sonnet
RICE, MVP scope, requirements
Market Researcher
Perplexity
TAM/SAM/SOM, competitive analysis, SWOT
UX Researcher
Sonnet
Personas, journeys, design tokens
Finance Analyst
Gemini
DCF, SaaS metrics, risk matrix, scenarios
Solutions Architect
Copilot
Feasibility, architecture, build-vs-buy TCO
Business Strategist
Gemini
GTM, business model, competitive moat
Domain Expert
Configurable
Industry SME, regulatory, domain risks
CEO / CTO Advisor
Opus
Go/No-Go verdict, kill switches, roadmap
Deliverables Writer
Sonnet
BRD, architecture, charter, risk register

Adaptive Cost Management

Budget-Adaptive Model Routing
60% → Opus downgrades 80% → Sonnet downgrades
$0 60% — Opus→Sonnet 80% — Sonnet→Haiku 100% — Block
CEO/CTO Advisor is always last to downgrade — stays on Opus until maximum budget pressure. Token estimation: (bytes ÷ 4) + 500 overhead per agent, tracked in state.json.

Quality Mode Presets

Quick Scout
Fast Filter
~$0.50–$1.50
PM + Architect + CEO. Light weight. Focused debate, 1 round. Perfect for early-stage filtering before you commit resources.
3 agents light mode 1 debate round budget-adaptive
Product Review
Feature Evaluation
~$2–$5
5 core experts + CEO. Deep analysis. Selective debate, 2 rounds. The default mode for shipping decisions and feature prioritization.
5 agents deep mode selective debate strict validation
Investment Grade
Major Decisions
~$8–$20
All 7–8 experts. Deep analysis. Relevance-based debate, 2 rounds. Multi-provider routing. For pivots, large bets, investor-ready analysis.
all agents multi-provider relevance debate full validation

v1.3.0 Features

⚙️
Role Controls
Enable/disable agents, set weight (light/deep), model (haiku/sonnet/opus), timeout per role in config.yaml.
🗣️
Selective Debate
4 debate modes: all-vs-all, selective pairs, relevance-based, focused topics. Max 1–2 rounds, configurable.
📁
Output Organization
Save evaluations by feature or project in organized folders: .gang/features/{name}/ or .gang/projects/{name}/.
💰
Cost Management
Token estimation, budget limits, warn_at 80%, block_at 100%. Per-model rates tracked in state.json.
🔀
Adaptive Model Routing
Budget-adaptive downgrade (opus→sonnet→haiku) + multi-provider: Perplexity, Gemini, GitHub Copilot.
🔍
Evidence Ledger
Codebase scan + web research populate evidence.json. Agents must cite evidence_ids for every major claim.
🧩
Assumptions Ledger
Agents register unstated hypotheses in assumptions.json with importance level and a validation plan.
📊
Scoring Rubrics
Rubric-anchored scores 1–10 with textual descriptions per level. Evidence and assumption linking required.
🔒
Advisor Guardrails
CEO/CTO can't issue GO without rubric-anchored scores. Auto-CONDITIONAL-GO on unvalidated assumptions.
✅
Validation Layer
Schema validation, cross-reference checks, between-stage validation, CI integration via validate-gang.sh.
🎛️
Quality Mode Presets
Quick Scout (~$1) · Product Review (~$5) · Investment Grade (~$20). Selected at /gang init, writes full config.

Commands

/gang runRun full pipeline end-to-end
/gang initScan project, select quality mode, populate evidence
/gang thinkCommittee setup + parallel expert analysis
/gang debateConfigurable cross-review debate
/gang scoreRubric-anchored plan synthesis and scoring
/gang adviseCEO/CTO advisory with guardrails
/gang deliverGenerate GO Package (requires GO/CONDITIONAL-GO)
/gang statusShow progress, committee, cost, validation
/gang configShow or edit configuration
/gang validateRun validation checks on all artifacts
/gang evaluationsList all feature and project evaluations