---
name: research-ai-topic
description: Research an AI topic using the so-yesterday.ai knowledge base via its REST API. Works without local repo access.
user-invocable: true
disable-model-invocation: true
allowed-tools: WebFetch
---

# Research AI Topic

Research any AI transformation topic using the so-yesterday.ai knowledge base. This skill works entirely through the public REST API — no local repository needed.

## Usage

`/research-ai-topic <topic>`

## Workflow

1. **Search** for the topic across all content types:
   ```
   GET https://so-yesterday.ai/api/search?q=<topic>&agent=true&limit=10
   ```
   Returns ranked results with snippets from videos, knowledge concepts, essays, digests, and personas.

2. **Read knowledge concepts** that match:
   ```
   GET https://so-yesterday.ai/api/knowledge/<slug>?agent=true
   ```
   Returns full markdown content with links rewritten as followable API URLs. Check the `related` field for connected concepts.

3. **Read video summaries** for deeper context:
   ```
   GET https://so-yesterday.ai/api/videos/<video_id>?agent=true
   ```
   Returns executive summary, key points, and full transcript.

4. **Explore the knowledge graph** for concept relationships:
   ```
   GET https://so-yesterday.ai/api/graph/data
   ```
   Returns nodes (concepts) and edges (relationships) as JSON.

5. **Synthesize** findings into a briefing with:
   - Key insights from knowledge concepts
   - Supporting evidence from video summaries
   - Links to source material (use `/knowledge/<slug>` and `/videos/<id>` paths)

## Output Format

Provide a structured briefing:
- **Summary** (3-5 sentences)
- **Key Concepts** (with links to knowledge pages)
- **Recent Developments** (from video summaries)
- **Related Topics** (from knowledge graph connections)