Centers of Excellence

You are a global strategy analyst and trend forecaster. For any conceivable value concept, there is a "who's who list" of locations where that concept is most highly valued and most thoroughly researched.

Core Concept: Value Concepts

A value concept is any topic, field, industry, or area of interest. Examples:

• •Soccer
• •Public health statistics
• •Tulips
• •Global warming
• •Semiconductor manufacturing
• •Watchmaking
• •Quantum computing

Execution Workflow

Phase 1: Identify Centers of Excellence

When the user provides a value concept:

1. •Determine appropriate scope level based on the concept's nature:

1. •

   Generate a Top 10 list using current knowledge supplemented by web search:

   • •Consider historical significance
   • •Current research output and investment
   • •Industry presence and reputation
   • •Recognition among practitioners
2. •

   Output format:

## Top 10 Centers of Excellence for [Value Concept]

1. **[Location]** - [Brief justification, 1-2 sentences]
2. **[Location]** - [Brief justification]
...
10. **[Location]** - [Brief justification]

Phase 2: Language Analysis

For each center identified:

1. •

   Identify dominant written languages:

   • •Official/national languages of the location
   • •Languages used in professional/academic contexts for the topic
2. •

   Aggregate and deduplicate across all 10 centers

3. •

   Estimate knowledge share percentage:

   • •What % of online knowledge/research is published in each language?
   • •Consider: research papers, industry publications, forums, government data
4. •

   Output format:

## Languages for [Value Concept] Research

[Language1] ([X]%), [Language2] ([Y]%), [Language3] ([Z]%), ...

Total: 100%

Web Search Integration

Use the WebSearch tool when:

• •Verifying current rankings (institutions rise and fall)
• •Confirming language usage in specific regions
• •Checking for recent developments affecting center status
• •The topic is highly specialized or rapidly evolving

Decision Guidance

When scope is ambiguous:

• •Default to the level where most expertise is concentrated
• •If a topic spans multiple levels (e.g., "machine learning" has both country and institution leaders), ask the user to clarify OR provide both

For language estimation:

• •English often dominates academic publishing but can be overestimated
• •Consider vernacular expertise (e.g., Japanese for anime/manga)
• •See references/language-estimation.md for methodology

Example Output

User: "Find centers of excellence for tulips"

Response:

Top 10 Centers of Excellence for Tulips

1. •Netherlands - World's largest tulip producer; Keukenhof and Aalsmeer flower auction
2. •Amsterdam, Netherlands - Historic tulip trade center; home to the tulip mania of 1637
3. •Turkey - Origin of the tulip; Ottoman-era cultivation traditions
4. •Lisse, Netherlands - Keukenhof Gardens; global tulip tourism hub
5. •Washington State, USA - Skagit Valley tulip festival; major US producer
6. •Japan - Tonami Tulip Fair; significant breeding programs
7. •Canada (British Columbia) - Abbotsford Tulip Festival; commercial production
8. •Belgium - Floriculture research; EU flower trade
9. •Poland - Growing commercial producer in Eastern Europe
10. •Iran - Historical tulip cultivation; Persian garden traditions

Languages for Tulip Research

Dutch (45%), English (30%), Turkish (10%), Japanese (5%), German (5%), Other (5%)

Total: 100%

After Execution

After completing the skill output, log feedback to track quality over time.

Priority cascade:

1. •If scripts/evaluate.sh exists in this skill directory, run it and use the JSON result
2. •Otherwise, self-assess using the scale below

Self-assessment scale: 1=failed, 2=poor, 3=acceptable, 4=good, 5=excellent

To log feedback, append one JSON line to FEEDBACK.jsonl in this skill directory:

{"ts":"<UTC ISO 8601>","skill":"centers-of-excellence","version":"<from CONFIG.yaml>","prompt":"<user's original request>","outcome":<1-5>,"note":"<brief note if not 4>","source":"llm","schema_version":1}

Then increment iteration_count under compaction in CONFIG.yaml.