Consulting Models

Multi-model consultation using cursor-agent CLI.

Operation Modes

1. CONSULTATION (Single Expert)

Ask a specific AI model for focused analysis.

Triggers: "Ask [model] about", "Consult [model] on", "What does [model] think"

Models: gpt, opus, gemini

Execute:

scripts/consult.sh "MODEL" "PROMPT"

Replace MODEL with alias, PROMPT with question. Output goes to stdout.

2. RESEARCH (Parallel Multi-Agent)

Launch multiple agents for comprehensive coverage.

Triggers: "Do research on", "Quick research:", "Extensive research on"

Intensities:

• •quick: 3 agents (1 per model)
• •standard: 9 agents (3 per model)
• •extensive: 24 agents (8 per model)

Execute:

scripts/research.sh "TOPIC" INTENSITY OUTPUT_DIR
scripts/collect.sh OUTPUT_DIR

See references/research-workflow.md for full workflow.

3. DELIBERATION (Multi-Round Debate)

Agents critique each other and refine positions over 4 rounds.

Triggers: "Deliberate on", "Have models debate", "Peer review:"

See references/deliberation-workflow.md for full protocol.

Quick Reference

# Single consultation
scripts/consult.sh "gpt" "What's the best rate limiting approach?"

# Parallel research
scripts/research.sh "API authentication" standard ./scratch/research
scripts/collect.sh ./scratch/research

Model Selection

Full details: references/models.md

Synthesis

After collecting results:

1. •Identify consensus (3+ models agree)
2. •Note disagreements
3. •Rate confidence (High/Medium/Low)
4. •Provide recommendation

Template: references/synthesis-template.md