inspirer

Think outside the box, backed by evidence. 12 provocation lenses, web-grounded research, scored and filtered recommendations.

Contents

• Quick Start
• Architecture
• Stage 1: FRAME
• Stage 2: DIVERGE
• Stage 3: RESEARCH
• Stage 4: SCORE
• Stage 5: CONVERGE
• Stage 6: DELIVER
• Depth Control
• Evolve-Loop Integration
• Output Schema
• Reference

Quick Start

# Basic — think creatively about a topic
/evo:inspirer "How should we handle real-time sync in a serverless app?"

# Control depth (QUICK = fast ideation, DEEP = thorough exploration)
/evo:inspirer "What features increase user retention?" --depth DEEP

# Specify number of lenses (default: 4 for STANDARD)
/evo:inspirer "Multi-agent coordination patterns" --lenses 5

# Output for evolve-loop consumption
/evo:inspirer "Improve eval infrastructure" --format evolve --depth QUICK

Parse arguments:

• First quoted string or remaining text → topic
• --depth QUICK|STANDARD|DEEP → research depth (default: STANDARD)
• --lenses N → number of provocation lenses (default: per depth level)
• --format full|brief|evolve → output format (default: full)

Architecture

Six-stage pipeline: FRAME → DIVERGE → RESEARCH → SCORE → CONVERGE → DELIVER

Input: User's topic/question
         │
    ┌────▼────┐
    │  FRAME  │  Parse topic, classify domain, detect constraints
    └────┬────┘
         │
    ┌────▼─────┐
    │ DIVERGE  │  Apply 3-5 provocation lenses → divergent questions
    └────┬─────┘
         │ 1-2 research questions per lens
    ┌────▼─────┐
    │ RESEARCH │  Web search each question (Smart or Default routing)
    └────┬─────┘
         │ Research findings with sources
    ┌────▼────┐
    │  SCORE  │  Create Inspiration Cards, score feasibility × impact × novelty
    └────┬────┘
         │ Scored cards with KEEP/DROP verdicts
    ┌────▼─────┐
    │ CONVERGE │  Rank by composite, diversity filter, select top 5-8
    └────┬─────┘
         │
    ┌────▼─────┐
    │ DELIVER  │  Output as Inspiration Report / brief table / evolve JSON
    └──────────┘