Kannaktopus Agent Catalog

32 specialized personas, infinite possibilities. But which one grabs YOUR problem? 🐙

Note on Permissions: Current releases prompt before background provider execution so you can approve Codex or Gemini usage explicitly.

TL;DR - Which Tentacle Do I Need?

Still unsure? Just describe what you need - Kannaktopus auto-routes to the right tentacle! 🦑

Knowledge-worker personas for researchers, consultants, and product managers are also included. See Knowledge Workers Guide →

Table of Contents

1. By Development Phase - Double Diamond workflow
2. By Domain - Backend, Frontend, Quality, Ops
3. Knowledge Work - Research, consulting, writing
4. Top 10 Tentacles - The heavy hitters
5. Cost Tiers - Token budget awareness
6. Common Workflows - Multi-tentacle patterns
7. The Principles (Don't Invoke Directly) - Quality checklists
8. Decision Tree - When all else fails

By Development Phase

Kannaktopus follows the Double Diamond methodology:

     DISCOVER         DEFINE         DEVELOP          DELIVER
      (probe)         (grasp)        (tangle)          (ink)

    \         /     \         /     \         /     \         /
     \  🔍  /       \  🎯  /       \  🔧  /       \  ✅  /
      \ 🦑 /         \    /         \ 🦑 /         \    /
       \  /           \  /           \  /           \  /
        \/             \/             \/             \/

   Diverge then      Converge to      Diverge with     Converge to
    converge          problem          solutions        delivery

🔍 Probe Phase: Research & Discovery

Use when: Starting a project, exploring options, gathering requirements

Pro tip: Probe tentacles are cheap (sonnet/inherit) - fire liberally!

🎯 Grasp Phase: Architecture & Design

Use when: Planning architecture, defining system design, creating blueprints

Pro tip: These tentacles are premium (opus/inherit) - they're worth it for architecture decisions.

🔧 Tangle Phase: Development & Implementation

Use when: Writing code, implementing features, building solutions

Pro tip: Tangle has a 75% quality gate - multiple tentacles validate before delivery.

✅ Ink Phase: Review & Delivery

Use when: Code review, security audit, performance optimization, shipping

Pro tip: Ink tentacles work with the squeeze command for adversarial review!

By Domain

Backend Development 🔌

General Backend    → backend-architect (inherit - adapts to complexity)
Database Focus     → database-architect (opus - critical decisions)
GraphQL API        → graphql-architect (opus - schema design)
Python Backend     → python-pro (opus - advanced patterns)
Node.js Backend    → typescript-pro (opus - type safety)

Frontend Development 🎨

React/Next.js      → frontend-developer (inherit - adapts)
TypeScript         → typescript-pro (opus - complex types)
Performance        → performance-engineer (inherit - profiling)

Quality & Security 🛡️

Code Quality       → code-reviewer (opus - comprehensive)
Security           → security-auditor (opus - critical)
Testing            → test-automator (sonnet - generation)
Performance        → performance-engineer (inherit - profiling)

Operations & Infrastructure ⚙️

Cloud/AWS/GCP      → cloud-architect (opus - critical decisions)
Kubernetes/Docker  → deployment-engineer (haiku - fast)
Troubleshooting    → devops-troubleshooter (sonnet - specialized)
Incidents          → incident-responder (sonnet - SRE)

Documentation & Diagrams 📚

Technical Docs     → docs-architect (sonnet - writing)
Flowcharts         → mermaid-expert (haiku - simple)

Knowledge Work

When your deliverable isn't code, but crystallized knowledge. 🎓

Kannaktopus also includes specialized agents for researchers, consultants, designers, marketers, and product managers.

🎯 UX Research

User interviews    → ux-researcher (opus - synthesis)
Journey mapping    → ux-researcher (opus - empathy)
Pain points        → ux-researcher (opus - definition)

📊 Strategy & Consulting

Market analysis    → strategy-analyst (opus - frameworks)
Business cases     → strategy-analyst (opus - quantification)
Exec summaries     → exec-communicator (sonnet - clarity)

📚 Academic & Research

Literature review  → research-synthesizer (opus - synthesis)
Gap analysis       → research-synthesizer (opus - identification)
Academic writing   → academic-writer (sonnet - formatting)

📋 Product Management

PRDs & specs       → product-writer (sonnet - clarity)
User stories       → product-writer (sonnet - structure)
Acceptance criteria→ product-writer (sonnet - precision)

📣 Marketing & Growth

Campaign planning  → marketing-strategist (sonnet - channels)
Growth funnels     → marketing-strategist (sonnet - optimization)
SEO & content      → marketing-strategist (sonnet - strategy)

💰 Finance & Pricing

Financial models   → finance-analyst (sonnet - projections)
Unit economics     → finance-analyst (sonnet - metrics)
Pricing strategy   → finance-analyst (sonnet - optimization)

🎨 UI/UX Design

Design systems     → ui-ux-designer (sonnet - BM25 intelligence)
Color & typography → ui-ux-designer (sonnet - data-driven)
Component specs    → ui-ux-designer (sonnet - handoff)

⚖️ Legal & Compliance

GDPR/CCPA review   → legal-compliance-advisor (sonnet - regulations)
SOC 2 readiness    → legal-compliance-advisor (sonnet - controls)
Contract review    → legal-compliance-advisor (sonnet - risk)

Deep dive: See the full Knowledge Workers Guide →

Top 10 Tentacles

The heavy hitters, based on real-world usage:

1. 🏗️ backend-architect

Model: inherit (adapts to complexity)

Superpower: API design, microservices architecture, distributed systems

When to use:

• Designing REST/GraphQL/gRPC APIs
• Architecting microservices systems
• Defining service boundaries and communication patterns
• Planning event-driven architectures

Avoid if:

• Simple CRUD operations (just implement directly)
• Frontend-only work (use frontend-developer)
• Database schema design (use database-architect first)

2. 🔍 code-reviewer

Model: opus (complex reasoning)

Superpower: Code quality, best practices, architecture review

When to use:

• PR reviews (pairs with /octo:review skill)
• Architecture validation
• Best practices enforcement

Avoid if:

• Security-focused review (use security-auditor)
• Performance-focused review (use performance-engineer)

3. 🐛 debugger

Model: sonnet (specialized analysis)

Superpower: Error analysis, stack traces, root cause identification

When to use:

• Failing tests or production errors
• Understanding cryptic stack traces
• Root cause analysis

Avoid if:

• Infrastructure issues (use devops-troubleshooter)

4. 🛡️ security-auditor

Model: opus (critical analysis)

Superpower: OWASP Top 10, vulnerability scanning, threat modeling

When to use:

• Security audits (pairs with squeeze command)
• Finding vulnerabilities in auth/payment code
• OWASP compliance checks

5. 🧪 tdd-orchestrator

Model: opus (multi-step reasoning)

Superpower: Test-driven development, red-green-refactor

When to use:

• Building new features test-first
• Comprehensive test coverage
• Refactoring with confidence

6. 🎨 frontend-developer

Model: inherit (adapts to complexity)

Superpower: React 19, Next.js 15, modern UI patterns

When to use:

• React component architecture
• Server Components and Client Components
• State management decisions

7. 🗃️ database-architect

Model: opus (critical decisions)

Superpower: Schema design, migrations, query optimization

When to use:

• New database schema design
• Migration planning
• Query performance issues

8. ⚡ performance-engineer

Model: inherit (adapts to profiling depth)

Superpower: Profiling, benchmarking, Core Web Vitals

When to use:

• N+1 query hunting
• API response time optimization
• Memory leak detection

9. 🐍 python-pro

Model: opus (advanced patterns)

Superpower: Python 3.12+, FastAPI, async programming

When to use:

• Python backend development
• FastAPI/Django/Flask services
• Async/await patterns

10. 📘 typescript-pro

Model: opus (complex type systems)

Superpower: Advanced types, generics, strict TypeScript

When to use:

• Advanced TypeScript type definitions
• Generic types and utility type creation
• Strict TypeScript configuration

Cost Tiers

💎 Premium (Opus)

Most expensive, use for critical decisions.

code-reviewer, tdd-orchestrator, database-architect, security-auditor, cloud-architect, python-pro, typescript-pro, graphql-architect

💰 Standard (Sonnet)

Mid-range, good for focused tasks.

debugger, test-automator, docs-architect, business-analyst, devops-troubleshooter, incident-responder, ui-ux-designer, marketing-strategist, finance-analyst, legal-compliance-advisor

💵 Trivial (Haiku)

Cheapest, use for simple tasks.

mermaid-expert, deployment-engineer

🔄 Adaptive (Inherit)

Auto-selects based on task complexity.

backend-architect, frontend-developer, ai-engineer, performance-engineer, context-manager

Common Workflows

🚀 New Feature Development

1. Research     → ai-engineer
2. Design       → backend-architect + database-architect
3. Implement    → tdd-orchestrator
4. Review       → code-reviewer + security-auditor

🔥 Production Bug Fix

1. Respond      → incident-responder
2. Debug        → debugger
3. Fix          → backend-architect
4. Validate     → test-automator

⚡ Performance Optimization

1. Profile      → performance-engineer
2. Database     → database-architect
3. Validate     → performance-engineer

🛡️ Security Audit

1. Scan         → security-auditor
2. Fix          → backend-architect
3. Validate     → security-auditor

The Principles (Internal Use Only)

⚠️ Don't invoke these directly! They're quality checklists used BY other tentacles.

Decision Tree

See agent-decision-tree.md for visual flowcharts.

Quick version:

What phase are you in?
├─ Research/Explore → PROBE tentacles
├─ Design/Plan     → GRASP tentacles
├─ Build/Code      → TANGLE tentacles
└─ Review/Ship     → INK tentacles

Quick Links

• README: ../README.md
• Decision Tree: agent-decision-tree.md
• Monthly Review: monthly-agent-review.md
• Personas: ../agents/personas/

🐙 "39 tentacles, one purpose: your code, done right." 🐙