AI-Agent-Toolkit debate

Debate Skill

install

source · Clone the upstream repo

git clone https://github.com/ngapngap/AI-Agent-Toolkit

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/ngapngap/AI-Agent-Toolkit "$T" && mkdir -p ~/.claude/skills && cp -r "$T/.agent/skills/debate" ~/.claude/skills/ngapngap-ai-agent-toolkit-debate && rm -rf "$T"

manifest: .agent/skills/debate/SKILL.md

source content

Debate Skill

Overview

Council-style debate skill for decision making with multiple perspectives.

Modes

IDE Mode (Single Model)

One model plays multiple roles sequentially.

Runner Mode (Multi-Model)

Multiple models participate as different council members.

Council Roles

Role	Responsibility
Architect A	Propose practical, proven solution
Architect B	Propose alternative approach
Critic C	Identify weaknesses and gaps
Devil's Advocate D	Challenge assumptions
Lead	Moderate discussion and make final decision

Scoring Rubric

Criterion	Weight	Description
constraint_compliance	0.30	Fits within project constraints
practicality	0.25	Can be implemented effectively
risk_awareness	0.25	Risks identified and mitigated
clarity	0.20	Clear and understandable

Usage

# Run debate in IDE mode (single model)
node .agent/skills/debate/scripts/debate.js --mode ide --input research.shortlist.json

# Run debate with config file
node .agent/skills/debate/scripts/debate.js --config debate-config.json

Inputs

```
20_research/research.shortlist.json
```

20_research/research.reuse_assessment.json

```
20_research/research.patterns.md
```

Outputs

```
30_debate/debate.inputs_for_spec.json
```
```
30_debate/debate.decision.md
```

Output Schema

See

schemas/debate.inputs_for_spec.schema.json

Decision Lock

Once decision is locked, changes require re-running Debate
Prevents silent scope changes during implementation