template.yaml

# Trinity-Compatible Agent Configuration
# Cornelius - The Insight Harvester & Second Brain Partner

# Required Fields
name: cornelius
display_name: "Cornelius"
description: |
  AI Insight Harvester and Second Brain Partner managing Eugene's Obsidian knowledge base
  (2,597 notes, 532 permanent notes, 174 AI insights, 1,590 document insights across 89 sessions).

  Core capabilities:
  - Problem-Solving & Advice: Ground advice in accumulated KB insights and personal frameworks
  - Insight Harvesting: Detect and capture original thinking, personal frameworks, and distinctive viewpoints
  - Knowledge Synthesis: Combine captured insights to create new content or discover patterns
  - Deep Research: Autonomous pipeline for discovering, extracting, and integrating cutting-edge research
  - Connection Discovery: Find non-obvious relationships between notes across domains
  - Dialectic Analysis: Stress-test ideas with committed positions and structural contradiction analysis
  - Content Generation: Create articles and perspectives grounded in knowledge base
  - Forecasting Integration: Access crowd predictions from Good Judgment Open

resources:
  cpu: "2"
  memory: "4g"

# Optional Metadata
version: "1.25.0"
author: "Eugene Yan"
updated: "2026-03-12"
tagline: "Your insight harvester, second brain partner, and KB-grounded advisor"

use_cases:
  - "Get KB-grounded advice on problems using your own frameworks"
  - "Quick knowledge lookup with semantic search and graph neighborhood"
  - "Extract unique insights from conversations and transcripts"
  - "Discover hidden connections between notes in your knowledge base"
  - "Generate articles and content from accumulated knowledge"
  - "Run autonomous deep research pipelines (discover, extract, connect)"
  - "Stress-test ideas with dialectic analysis (pro/con agents + synthesis)"
  - "Query crowd forecasts from Good Judgment Open"
  - "Analyze Economist articles and integrate insights"
  - "Graduate notes through Zettelkasten workflow"
  - "Search and recall relevant knowledge using 3-layer semantic search"
  - "Analyze and understand your knowledge base structure"

capabilities:
  - semantic_search
  - spreading_activation
  - usage_based_learning
  - insight_extraction
  - connection_discovery
  - knowledge_synthesis
  - content_generation
  - problem_solving
  - dialectic_analysis
  - deep_research
  - forecasting_integration
  - current_affairs_analysis
  - belief_system_management
  - epub_processing
  - diagram_generation
  - image_generation

# Credential Schema
credentials:
  mcp_servers:
    aistudio:
      env_vars:
        - GEMINI_API_KEY
    trinity:
      env_vars:
        - TRINITY_API_KEY
  env_file:
    - VAULT_BASE_PATH
    - DOCUMENT_INSIGHTS_PATH
    - TRINITY_REMOTE_AGENT

# Sub-Agents
sub_agents:
  - name: vault-manager
    description: "CRUD operations on Obsidian vault notes with knowledge graph integrity"
  - name: connection-finder
    description: "Discover hidden connections and relationships between permanent notes"
  - name: auto-discovery
    description: "Autonomous cross-domain connection hunter using analytical reasoning"
  - name: insight-extractor
    description: "Extract unique insights from personal content, preserving authentic voice"
  - name: document-insight-extractor
    description: "Extract insights from external research (papers, books, articles)"
  - name: thinking-partner
    description: "Brainstorming and ideation support through dialogue"
  - name: diagram-generator
    description: "Create Mermaid diagrams and visualizations"
  - name: local-brain-search
    description: "SYNAPSE-inspired vector search with spreading activation and usage-based learning"
  - name: research-specialist
    description: "Deep web research with Gemini AI, Google Search grounding, and Apollo.io"

# Skills (Slash Commands)
skills:
  # Problem-Solving & Advice
  - name: advise
    description: "KB-grounded problem-solving with framework application and citations"
  - name: dialectic
    description: "Stress-test ideas with two committed positions and structural synthesis"

  # Fast Knowledge Access
  - name: quick-search
    description: "Fast KB lookup (3 tool calls max) - search + graph + read"
  - name: recall
    description: "3-layer semantic search with spreading activation"
  - name: search-vault
    description: "Quick search combining semantic and keyword approaches"

  # Connection & Discovery
  - name: find-connections
    description: "Discover hidden connections around a note or topic"
  - name: auto-discovery
    description: "Autonomous cross-domain connection hunting"
  - name: integrate-recent-notes
    description: "Find and connect notes created in the last 14 days"

  # Research & Extraction
  - name: deep-research
    description: "Autonomous research pipeline - discover, extract, connect"
  - name: extract-insights
    description: "Extract unique insights from personal content"
  - name: extract-document-insights
    description: "Extract insights from external research documents"
  - name: economist-analyze
    description: "Scrape Economist articles, extract insights, discover connections"

  # Content Creation
  - name: get-perspective-on
    description: "Extract unique perspective on a topic (used by Ruby)"
  - name: create-article
    description: "Generate article from KB insights with tone/structure guides"
  - name: synthesize-insights
    description: "Combine multiple insights into coherent narrative"

  # Forecasting
  - name: gjopen-lookup
    description: "Query GJ Open forecasts for crowd predictions"
  - name: gjopen-refresh
    description: "Update forecasting data from Good Judgment Open"
  - name: gjopen-heartbeat
    description: "Full Playwright scrape + git sync for GJ Open"

  # Knowledge Management
  - name: graduate-insights
    description: "Review and graduate notes to permanent status using Zettelkasten"
  - name: analyze-kb
    description: "Analyze knowledge base structure and generate report"
  - name: refresh-index
    description: "Rebuild Local Brain Search FAISS index"
  - name: update-changelog
    description: "Update the master CHANGELOG.md"

  # Utility
  - name: talk
    description: "Interactive conversational mode for brainstorming"
  - name: nano-banana-image-generator
    description: "Generate images using Gemini 2.5 Flash ($0.039/image)"
  - name: epub-chapter-extractor
    description: "Extract chapters from EPUB files to markdown"
  - name: self-diagnostic
    description: "Run agent health check on skills, tools, integrations"

  # Trinity Integration
  - name: trinity-schedules
    description: "Manage scheduled tasks on Trinity platform"
  - name: git-commit-push
    description: "Stage, commit, and push changes with approval gate"

# MCP Servers
mcp_servers:
  - name: mermaid-diagram
    description: "Generate Mermaid diagrams (PNG/SVG) from markdown"
  - name: aistudio
    description: "Gemini AI for content generation and Google search grounding"
  - name: ebook-mcp
    description: "EPUB and PDF processing for ebook content extraction"
  - name: playwright
    description: "Browser automation for web scraping (local only)"
  - name: trinity
    description: "Remote agent orchestration and scheduling"

# Git Configuration
git:
  push_enabled: true
  commit_paths:
    - memory/
    - plans/
    - Brain/
    - outputs/
    - resources/gjopen-questions.json
  ignore_paths:
    - .mcp.json
    - .env
    - .trinity/
    - .claude/commands/trinity/
    - resources/local-brain-search/data/

# Planning Configuration
planning:
  enabled: true
  max_active_plans: 5
  default_task_timeout_minutes: 60
  auto_checkpoint: true

# Shared Folders (for inter-agent collaboration)
shared_folders:
  expose: true
  consume: true

# Custom Metrics
metrics:
  - name: notes_total
    type: gauge
    label: "Total Notes"
    description: "Total number of notes in the knowledge base"
    unit: "notes"
  - name: permanent_notes
    type: gauge
    label: "Permanent Notes"
    description: "Core atomic insights (02-Permanent + AI Extracted)"
    unit: "notes"
  - name: document_insights
    type: gauge
    label: "Document Insights"
    description: "Research session extractions"
    unit: "notes"
  - name: insights_extracted
    type: counter
    label: "Insights Extracted"
    description: "Number of insights extracted this session"
    unit: "insights"
  - name: connections_discovered
    type: counter
    label: "Connections Found"
    description: "Number of connections discovered this session"
    unit: "connections"
  - name: search_queries
    type: counter
    label: "Search Queries"
    description: "Number of semantic searches performed"
    unit: "queries"
  - name: advice_requests
    type: counter
    label: "Advice Requests"
    description: "Number of /advise queries handled"
    unit: "requests"
  - name: index_status
    type: status
    label: "Index Status"
    description: "Status of the local FAISS index"
    values:
      - value: "indexed"
        color: "green"
        label: "Indexed"
      - value: "stale"
        color: "yellow"
        label: "Needs Reindex"
      - value: "missing"
        color: "red"
        label: "Not Indexed"
  - name: memory_learning
    type: status
    label: "Memory Learning"
    description: "Status of usage-based learning system"
    values:
      - value: "active"
        color: "green"
        label: "Learning"
      - value: "disabled"
        color: "gray"
        label: "Disabled"