@rune-kit/rune 2.3.3 → 2.4.0

package/docs/references/claudekit-analysis.md

@@ -1,414 +0,0 @@
-# ClaudeKit Competitive Analysis (Deep Research)
-
-> Reference document. Not part of Rune's public docs.
-> Date: 2026-02-27 | Updated: 2026-03-10 (deep research)
-
----
-
-## What ClaudeKit Is
-
-ClaudeKit ($99-149) sells pre-built skills, workflows, and agents for Claude Code.
-Built by mrgoonie (18+ year CTO), battle-tested in real products, 4,000+ users across 109 countries.
-
-- **Total**: 108+ skills, 95+ commands, 45 AI agents
-- Engineer Kit: $99 — 70+ skills, 40+ workflows, 17 agents, 76 slash commands
-- Marketing Kit: $99 — 8 MCP integrations, 32 agents, 68 skills, 119 commands, 5 workflows
-- Bundle: $149 — both kits, lifetime access, 14-day money-back guarantee
-- Payment: Stripe → private GitHub repo access
-- Referral program: 20% recurring commission
-- CLI: `bun add -g claudekit-cli` then `ck new --dir my-project --kit engineer`
-- Also available as Claude Code Plugin Marketplace install
-
-## Distribution Model
-
-- Copies `.claude/` directory into each project (project-level, not global)
-- Skills also available via Plugin Marketplace: `/plugin install ai-ml-tools@claudekit-skills`
-- Requires `claude --dangerously-skip-permissions` for autonomous workflows
-- Central config: `.claude/CLAUDE.md` controls all agent behaviors, skills, workflows
-- Skills stored in `.claude/skills/` with auto-activation via codebase detection
-
-## Repository Structure
-
-```
-.claude/
-├── agents/          # 17 specialized agents (engineer) or 32 (marketing)
-├── commands/        # 35+ workflow automation commands
-├── skills/          # 40+ pre-built agent skills
-├── workflows/       # Workflow coordination files
-├── CLAUDE.md        # Central project context template
-├── docs/            # Production documentation
-└── plans/           # Implementation planning
-```
-
----
-
-## Engineer Kit — Complete Agent Roster (17)
-
-### Development & Implementation
-1. **planner** — Research, analyze, create implementation plans before coding
-2. **fullstack-developer** — Full-stack implementation with strict file ownership + parallel execution
-3. **researcher** — Parallel research & best practices discovery
-4. **database-admin** — Schema design & migrations
-
-### Quality & Testing
-5. **tester** — Comprehensive test suite generation, coverage analysis
-6. **code-reviewer** — Security audits, performance analysis, code quality
-7. **code-simplifier** — Autonomous code refinement for clarity and maintainability
-8. **debugger** — Root cause analysis, log investigation, issue diagnosis
-
-### Design & Creative
-9. **ui-ux-designer** — Screenshot to production UI + visual asset generation (Three.js, responsive)
-10. **brainstormer** — Creative brainstorming, challenge assumptions, debate decisions
-11. **copywriter** — Marketing copy & changelogs
-
-### Documentation & Management
-12. **docs-manager** — Technical documentation, API docs, architecture guides (auto-updating)
-13. **project-manager** — Progress tracking, cross-agent coordination, status reports
-14. **journal-writer** — Document failures and setbacks with brutal honesty
-15. **git-manager** — Conventional commits, security scanning, token-optimized
-
-### Integration & Research
-16. **scout** — Codebase exploration & discovery
-17. **scout-external** — External tools integration & discovery
-
-### Agent Orchestration Patterns
-- **Sequential**: planner → developer → tester → reviewer → git-manager
-- **Parallel**: Independent agents work simultaneously, aggregate results
-- **Hybrid**: Combines both for complex tasks
-- Communication via shared files, handoff protocols, real-time progress tracking
-
----
-
-## Marketing Kit — Complete Agent Roster (32)
-
-### TOFU (Top of Funnel — Attraction) — 4 agents
-- **attraction-specialist** — Keyword research, competitor content intelligence, landing page gen, programmatic SEO
-- **seo-specialist** — Technical audits, content optimization, Search Console monitoring
-- **lead-qualifier** — Lead scoring and intent detection
-- **researcher** — Market intelligence and competitive trends
-
-### MOFU (Middle of Funnel — Nurture) — 5 agents
-- **email-wizard** — Sequence generation, dynamic personalization, send-time optimization, A/B testing
-- **sale-enabler** — Sales collateral and pitch materials
-- **funnel-architect** — Conversion optimization
-- **content-creator** — Multi-format gen (blogs, videos, ad content), brand consistency, editorial calendars
-- **continuity-specialist** — Customer retention and engagement
-
-### BOFU (Bottom of Funnel — Convert) — 1 agent
-- **upsell-maximizer** — Revenue expansion, product recommendations, revenue forecasting, feature adoption tracking
-
-### Core Operations — 5 agents
-- **campaign-manager** — Multi-channel coordination, budget optimization, performance tracking, launch workflows
-- **copywriter** — High-converting copywriting
-- **brainstormer** — Strategy and ideation
-- **content-reviewer** — Quality control and brand voice consistency
-- **campaign-debugger** — Performance issue diagnosis
-
-### Community & Social — 2 agents
-- **social-media-manager** — Multi-platform content management
-- **community-manager** — Engagement and moderation
-
-### Support & Infrastructure — 15 agents
-- planner, project-manager, docs-manager, git-manager, journal-writer
-- scout, scout-external, mcp-manager, analytics-analyst, tester
-- (5 additional unnamed support agents)
-
-### 8 MCP Integrations
-1. **Google Analytics 4** — Analytics
-2. **Google Search Console** — SEO monitoring
-3. **Google Ads** — Ads management
-4. **SendGrid** — Email campaigns
-5. **Resend** — Email campaigns (alternative)
-6. **Slack** — Team communication
-7. **Discord** — Community management
-8. **ReviewWeb** — Review aggregation
-
-### 5 Marketing Workflows
-1. **Marketing Workflow** — Research → Insights → Creative → Plan → Create → Edit → Publish → Measure
-2. **Sales Workflow** — Lead gen → Qualification → Nurture → Close
-3. **CRM Workflow** — Contact management + automation
-4. **Video Production Workflow** — Script → Record → Edit → Publish
-5. **Design Workflow** — Brief → Concept → Create → Review
-
----
-
-## Complete Skills Catalog (from GitHub: mrgoonie/claudekit-skills)
-
-### Authentication & Security
-- **better-auth** — TypeScript auth framework (email/password, OAuth, 2FA, passkeys, multi-tenancy)
-
-### AI & Machine Learning
-- **ai-multimodal** — Google Gemini API (audio transcription, image analysis, video processing, doc extraction, image gen)
-- **context-engineering** — AI agent optimization, degradation patterns, memory architectures
-- **google-adk-python** — Google Agent Development Kit for multi-agent orchestration
-
-### Backend Development
-- **backend-development** — Node.js, Python, Go, Rust + NestJS, FastAPI, Django frameworks
-
-### Frontend & Web
-- **web-frameworks** — Next.js, Turborepo, RemixIcon
-- **ui-styling** — shadcn/ui, Radix UI, Tailwind CSS, accessible components
-- **frontend-design** — Production-grade interface creation
-- **frontend-development** — React/TypeScript, Suspense, TanStack Router, MUI v7
-- **threejs** — 3D web (WebGL/WebGPU, animations, shaders)
-- **aesthetic** — Interface design principles (visual hierarchy, color theory, micro-interactions)
-
-### Browser Automation & Testing
-- **chrome-devtools** — Puppeteer CLI (automation, screenshots, performance analysis)
-- **web-testing** — Playwright, Vitest, k6 (E2E, unit, integration, load, security, visual, accessibility)
-
-### Cloud & DevOps
-- **devops** — Cloudflare (Workers, R2, D1, KV, Pages, Durable Objects), Docker, GCP (Compute, GKE, Cloud Run)
-
-### Databases
-- **databases** — MongoDB (aggregation, Atlas) + PostgreSQL (schema, queries, optimization)
-
-### Development Tools
-- **claude-code** — Claude Code features (slash commands, hooks, plugins, MCP servers)
-- **mcp-builder** — Build MCP servers (FastMCP Python / TypeScript)
-- **mcp-management** — Manage MCP servers and tools
-- **repomix** — Package repos into single AI-friendly files
-- **media-processing** — FFmpeg + ImageMagick for multimedia
-
-### Documentation & Research
-- **docs-seeker** — Documentation discovery (llms.txt, GitHub analysis)
-
-### Code Quality
-- **code-review** — Feedback evaluation + completion verification
-- **skill-creator** — Create reusable AI skills (SKILL.md < 100 lines, progressive disclosure)
-
-### Debugging (4 sub-skills)
-- **defense-in-depth** — Layer validation to prevent bugs structurally
-- **root-cause-tracing** — Backward tracing through call stacks
-- **systematic-debugging** — Four-phase debugging framework (reproduce → isolate → trace → fix)
-- **verification-before-completion** — Command verification before success claims
-
-### Document Processing (4 sub-skills)
-- **docx** — Word document creation with tracked changes
-- **pdf** — PDF extraction, creation, merging, form filling
-- **pptx** — PowerPoint creation and editing
-- **xlsx** — Spreadsheet building with formulas and financial modeling
-
-### E-commerce
-- **shopify** — Shopify apps, extensions, themes (GraphQL/REST)
-
-### Payments
-- **payment-integration** — SePay, Polar, Stripe, Paddle, Creem.io + webhook verification
-
-### Problem-Solving (6 sub-skills)
-- **collision-zone-thinking** — Combining unrelated concepts
-- **inversion-exercise** — Testing opposite assumptions
-- **meta-pattern-recognition** — Identifying universal principles
-- **scale-game** — Testing at extreme scales
-- **simplification-cascades** — Eliminating components via single insights
-- **when-stuck** — Dispatching to appropriate techniques
-
-### Advanced Reasoning
-- **sequential-thinking** — Step-by-step reasoning with revision
-
-### Visualization
-- **mermaidjs-v11** — 24+ diagram types (flowcharts, sequence diagrams, etc.)
-
-### Meta
-- **mcp-manager** — Subagent for managing MCP context without bloat
-
-**Total confirmed individual skills: ~40+ with sub-skills bringing count to ~55+**
-
----
-
-## Complete Slash Commands
-
-### Core Development
-- `/ck:plan <description>` — Create implementation plan
-- `/ck:plan --two` — Two-step implementation plan
-- `/ck:cook <description>` — Implement new features
-- `/ck:cook --auto <description>` — Autonomous feature implementation
-- `/ck:bootstrap <idea>` — Bootstrap new project (interactive)
-- `/ck:bootstrap --auto <idea>` — Fully autonomous project bootstrap
-- `/ck:test` — Run test suite and report
-- `/ck:debug` — Log analysis and root cause diagnosis
-- `/ck:scout` — Code analysis agent
-- `/ck:fix --quick <desc>` — Quick bug fix
-- `/ck:fix <description>` — Auto-detect complexity bug fix
-- `/ck:fix` (no args) — Auto-fetch logs or detect failing tests
-- `/ck:fix <github-action-url>` — Fix CI/CD pipeline issues
-
-### Documentation
-- `/ck:docs init` — Initialize documentation and specs
-- `/ck:docs update` — Update existing documentation
-- `/ck:docs summarize` — Summarize documentation
-
-### Git Operations
-- `/ck:git cm` — Create commit with meaningful message
-- `/ck:git cp` — Commit and push
-- `/ck:git pr` — Create pull request
-
-### Code Quality
-- `/review:codebase` — Full codebase review (security, performance, standards)
-- `/ck:brainstorm <description>` — Brainstorm technical approaches
-
-### Integration
-- `/integrate:polar` — Polar API integration
-- `/integrate:sepay` — SePay payment integration
-- `/integrate <service>` — Generic integration
-
-### Marketing-Specific
-- `/plan` — Campaign planning
-- `/content/good` — Landing page generation
-- `/content/cro` — CRO-optimized copy
-- `/seo:audit` — SEO audit
-- `/seo:keywords` — Keyword research and competitor analysis
-- `/campaign:email` — Email campaigns via SendGrid/Resend
-
-### Context Management
-- `/clear` — Context window reset
-
-**Total: 76+ engineer commands + 119 marketing commands = ~195 total**
-
----
-
-## Key Workflow: /bootstrap (Detailed)
-
-```
-/bootstrap "build a REST API with auth"
-→ Q&A: app type (REST/Web/Mobile/CLI), tech stack, DB, auth method
-→ Phase 1: Researcher agents find best practices, libraries, security considerations
-→ Phase 2: Planner agents create architecture, file structure, implementation steps, test strategy
-→ Phase 3: Create project structure
-→ Phase 4: Implement code (auth, CRUD, validation, rate limiting)
-→ Phase 5: Generate tests (87% coverage target)
-→ Phase 6: Generate docs (Swagger, README, architecture)
-→ Total: 5-8 minutes, fully autonomous
-```
-
-Autonomous mode: `/bootstrap:auto [detailed description]` — skips Q&A, requires detailed input.
-
-Output structure:
-```
-project/
-├── .claude/         # Custom commands, agent definitions, workflows
-├── src/             # Routes, models, middleware, utilities, entry point
-├── tests/           # Unit, integration, E2E
-└── docs/            # API docs, code standards, system architecture, codebase summary
-```
-
----
-
-## CCS (Claude Code Switch) — Multi-Account Tool
-
-- **Purpose**: Switch between multiple Claude accounts + AI models instantly
-- **Auth**: Browser-based OAuth (30 seconds, authenticate once)
-- **Profiles**: Each profile has own todos, sessions, logs (true isolation)
-- **Models**: Claude, Gemini 2.5 Pro, GPT-4o, o1, Antigravity via OpenRouter (300+ models)
-- **Cost savings**: $500-1000/month claimed — GLM handles 70% of tasks at 80% lower cost
-- **Dashboard**: React 19 real-time UI for profile management
-- **Website**: ccs.kaitran.ca
-- **Architecture**: v3.0 login-per-profile, CLIProxyAPI OAuth proxy
-
----
-
-## Architecture Deep Dive
-
-### Three Interconnected Systems
-1. **Agents** — Specialized AI assistants with defined roles
-2. **Commands/Skills** — Slash commands that invoke agents + knowledge modules
-3. **Workflows** — Coordination patterns ensuring agents work cohesively
-
-### Skill Mechanism
-- Files in `.claude/skills/skill-name/SKILL.md` (< 100 lines, imperative form)
-- Auto-activated via codebase detection (e.g., `next.config.js` → Next.js skill)
-- Progressive disclosure: core SKILL.md + `/scripts/`, `/references/`, `/assets/`
-- Distributable as ZIP packages
-
-### Agent Communication
-- Shared files for context passing (plans, code, test results)
-- Handoff protocols between agents
-- Real-time progress tracking visible to users
-- Sequential, Parallel, or Hybrid orchestration
-
-### Workflow Example: `/ck:cook "add auth with Better Auth"`
-1. System detects Better Auth skill → injects context
-2. Planner agent generates plan referencing skill patterns
-3. Developer agent writes code following skill guidelines
-4. Tester agent creates tests from skill examples
-5. Results consolidated for user
-
----
-
-## User Reviews & Testimonials
-
-- "Worth the $30/month, big time!" (pricing discrepancy — possibly old pricing or subscription model)
-- "It fixed all the CodeRabbit issues it found" — code review automation
-- "The /plan and /cook features are very impressive"
-- "Project alignment achieved within 1 session"
-- "Completed a feature with just a single prompt"
-- "Productivity doubled from 8 to 16 hours"
-- Users praise stability improvements over raw Claude Code
-- 4,000+ users across 109 countries
-
-## Rune vs ClaudeKit
-
-| Dimension | ClaudeKit | Rune |
-|-----------|-----------|------|
-| Install | Copy files per project | Global plugin |
-| Pricing | $99-149 one-time | Free (MIT) |
-| Automation | Fully autonomous (skip-permissions) | HARD-GATEs, user approves plan |
-| Philosophy | Hands-off, AI does everything | Developer stays in the loop |
-| Safety | Low (skips all permissions) | High (sentinel, gates, reviews) |
-| Mesh | No documented skill connections | 160+ bidirectional connections |
-| Cross-session | Not documented | .rune/ persistence |
-| Security | Not documented | sentinel, sast, integrity-check |
-| Skills count | 108+ skills, 95+ shortcuts, 45+ agents | 49 core + 12 L4 packs (free) + Pro packs (planned) |
-
-## Key Takeaways
-
-1. **Market validates demand** — people pay $99-149 for Claude Code productivity tools
-2. **Autonomous execution sells** — "describe and get result" is the primary value prop
-3. **No marketplace payment needed** — Stripe + private GitHub repo is the pattern
-4. **Project-level > plugin-level for some use cases** — copying .claude/ gives full control
-5. **`--dangerously-skip-permissions` is the enabler** — without it, autonomous workflows aren't possible
-6. **Marketing Kit is unique** — no equivalent in Rune (and shouldn't be per VISION.md anti-goals)
-
-## What Rune Should NOT Copy
-
-- Skip-permissions approach — contradicts Rune's safety philosophy
-- Per-project file copying — Rune's plugin model is cleaner
-- Quantity-first approach (108+ skills) — Rune optimizes for connections, not count
-- ~~Marketing/non-dev skills~~ — **REVISED**: Rune Pro will cover business departments as paid packs (separate repo)
-
-## Competitive Positioning (Updated 2025-03-10)
-
-### Numbers comparison
-
-| Metric | ClaudeKit | Rune Free | Rune Free + Pro (planned) |
-|--------|-----------|-----------|---------------------------|
-| Skills | 108+ | 49 core | 49 + ~54 pro sub-skills |
-| Shortcuts/Commands | 95+ | ~20 | ~65 (with pro workflows) |
-| Agents | 45+ | 21 (L3 utilities) | ~30 |
-| L4 Packs | N/A | 12 | 12 free + 9 pro |
-| Price | $99-149 | FREE | $29-49/mo or $199 lifetime |
-| Mesh connections | Not documented | 170+ | 200+ (pro adds cross-domain) |
-| Platforms | Claude Code only | 5 platforms | 5 platforms |
-| Safety | Skip-permissions | HARD-GATEs + sentinel | HARD-GATEs + sentinel |
-
-### Rune's differentiation
-
-1. **Free core with deep mesh** — ClaudeKit charges $99 for what Rune gives free
-2. **Multi-platform** — ClaudeKit is Claude Code only. Rune works on Cursor, Windsurf, Antigravity
-3. **Safety-first** — ClaudeKit requires `--dangerously-skip-permissions`. Rune uses HARD-GATEs
-4. **Mesh architecture** — 170+ connections vs flat skill list. Skills compose, not just coexist
-5. **Pro = business expansion** — ClaudeKit's Marketing Kit = $99 for 28 agents. Rune Pro = $29-49/mo for 9 packs across all departments
-
-### Where ClaudeKit wins
-
-1. **Autonomous execution** — `/bootstrap` generates full project in 5-8 min without user interaction
-2. **Quantity perception** — "108+ skills" sounds bigger than "49 skills" even if Rune's are deeper
-3. **Established market** — already has paying customers, proven demand
-4. **CCS multi-account** — account switching between providers (unique feature)
-
-## What Rune Could Learn (for other projects)
-
-- Autonomous `/bootstrap` workflow — powerful for greenfield projects
-- Interactive config before execution — better UX than "approve each phase"
-- Agent Teams orchestration — similar to Rune's `team` but more hands-off
-- One-command project generation — high perceived value

package/docs/references/voltagent-analysis.md

@@ -1,189 +0,0 @@
-# VoltAgent/awesome-claude-code-subagents Analysis
-
-> Reference document. Not part of Rune's public docs.
-> Date: 2026-03-11 | Source: https://github.com/VoltAgent/awesome-claude-code-subagents
-
----
-
-## Overview
-
-- **Stars**: 13,285 | **Forks**: 1,453
-- **Total agents**: 131 (.md files) across 10 categories
-- **License**: MIT
-- **Format**: Claude Code native subagent .md with YAML frontmatter
-- **Distribution**: Plugin marketplace, manual install, installer script, meta agent-installer
-
-## Agent File Structure
-
-```yaml
----
-name: kebab-case-name
-description: "When to invoke this agent"
-tools: Read, Write, Edit, Bash, Glob, Grep
-model: sonnet | opus | haiku | inherit
----
-
-You are a senior [role] with expertise in [domains]...
-
-When invoked:
-1. Query context manager for [X]
-2. Review [Y]
-3. Analyze [Z]
-4. Implement [W]
-
-## Communication Protocol
-### [Context] Assessment
-JSON-formatted context request
-
-## Development Workflow
-### 1. Analysis Phase
-### 2. Implementation Phase (with JSON progress tracking)
-### 3. Excellence Phase (delivery notification)
-
-Integration with other agents:
-- Collaborate with X on Y
-```
-
-### Key Fields
-
-| Field | Purpose | Values |
-|-------|---------|--------|
-| `name` | Agent identifier | kebab-case |
-| `description` | Auto-invocation trigger | Natural language |
-| `tools` | Permission scope | Comma-separated Claude Code tools |
-| `model` | Model routing | opus, sonnet, haiku, inherit |
-
-### Tool Assignment Philosophy
-
-| Role Type | Tools | Mode |
-|-----------|-------|------|
-| Read-only (reviewers, auditors) | Read, Grep, Glob | Read-only |
-| Research (analysts) | Read, Grep, Glob, WebFetch, WebSearch | Read-only |
-| Code writers (developers) | Read, Write, Edit, Bash, Glob, Grep | Read-write |
-| Documentation | Read, Write, Edit, Glob, Grep, WebFetch, WebSearch | Read-write |
-
-### Model Distribution
-
-| Model | Usage | ~% |
-|-------|-------|-----|
-| sonnet | Everyday coding, debugging, refactoring | 70% |
-| opus | Deep reasoning, architecture, security, finance | 20% |
-| haiku | Quick tasks, docs, search, dependency checks | 10% |
-
----
-
-## Full Agent Inventory (131 agents)
-
-### 01: Core Development (10)
-api-designer, backend-developer, electron-pro, frontend-developer, fullstack-developer, graphql-architect, microservices-architect (opus), mobile-developer, ui-designer, websocket-engineer
-
-### 02: Language Specialists (26)
-angular-architect, cpp-pro, csharp-developer, django-developer, dotnet-core-expert, dotnet-framework-4.8-expert, elixir-expert, flutter-expert, golang-pro, java-architect (opus), javascript-pro, kotlin-specialist, laravel-specialist, nextjs-developer, php-pro, powershell-5.1-expert, powershell-7-expert, python-pro, rails-expert, react-specialist, rust-engineer, spring-boot-engineer, sql-pro, swift-expert, typescript-pro, vue-expert
-
-### 03: Infrastructure (16)
-azure-infra-engineer, cloud-architect (opus), database-administrator, deployment-engineer, devops-engineer, devops-incident-responder, docker-expert, incident-responder, kubernetes-specialist, network-engineer, platform-engineer, security-engineer (opus), sre-engineer, terraform-engineer, terragrunt-expert, windows-infra-admin
-
-### 04: Quality & Security (14)
-accessibility-tester, ad-security-reviewer (opus), architect-reviewer (opus), chaos-engineer, code-reviewer (opus), compliance-auditor (opus), debugger, error-detective, penetration-tester (opus), performance-engineer, powershell-security-hardening, qa-expert, security-auditor (opus), test-automator
-
-### 05: Data & AI (12)
-ai-engineer (opus), data-analyst, data-engineer, data-scientist, database-optimizer, llm-architect (opus), machine-learning-engineer, ml-engineer, mlops-engineer, nlp-engineer, postgres-pro, prompt-engineer
-
-### 06: Developer Experience (13)
-build-engineer (haiku), cli-developer, dependency-manager (haiku), documentation-engineer (haiku), dx-optimizer, git-workflow-manager, legacy-modernizer, mcp-developer, powershell-module-architect, powershell-ui-architect, refactoring-specialist, slack-expert, tooling-engineer
-
-### 07: Specialized Domains (12)
-api-documenter (haiku), blockchain-developer, embedded-systems, fintech-engineer (opus), game-developer, iot-engineer, m365-admin, mobile-app-developer, payment-integration, quant-analyst (opus), risk-manager, seo-specialist (haiku)
-
-### 08: Business & Product (11)
-business-analyst, content-marketer (haiku), customer-success-manager (haiku), legal-advisor, product-manager (haiku), project-manager (haiku), sales-engineer (haiku), scrum-master (haiku), technical-writer (haiku), ux-researcher, wordpress-master
-
-### 09: Meta & Orchestration (10)
-agent-installer (haiku), agent-organizer, context-manager, error-coordinator, it-ops-orchestrator, knowledge-synthesizer, multi-agent-coordinator (opus), performance-monitor, task-distributor, workflow-orchestrator
-
-### 10: Research & Analysis (7)
-competitive-analyst (haiku), data-researcher (haiku), market-researcher (haiku), research-analyst (haiku), scientific-literature-researcher (haiku), search-specialist (haiku), trend-analyst (haiku)
-
----
-
-## Quality Assessment
-
-### Strengths
-- Clean, consistent template across all 131 agents
-- Tool permission separation (read-only vs write)
-- Model routing per agent (haiku/sonnet/opus/inherit)
-- Good distribution UX (installer, catalog search)
-- Broad domain coverage
-
-### Weaknesses
-- **No decision logic** — agents are topic outlines, not workflows
-- **No inter-agent connections** — "Collaborate with X" is passive suggestion, not enforced
-- **Zero personality** — all agents use identical template/voice
-- **Duplicate agents** — ml-engineer vs machine-learning-engineer, mobile-developer vs mobile-app-developer
-- **Fake example metrics** in templates (234K messages/min, $4.2M revenue)
-- **No context management infrastructure** — relies on native Claude Code only
-- **No escalation chains** — if agent fails, no fallback routing
-- **Claude Code only** — no multi-platform
-
----
-
-## Rune vs VoltAgent
-
-| Dimension | VoltAgent (131) | Rune (55 + 72 in L4 = ~127) |
-|-----------|----------------|------------------------------|
-| Architecture | Flat catalog, 10 categories | 5-layer mesh, enforced delegation |
-| Inter-agent connections | Passive suggestions | 200+ enforced + trade matrix |
-| Routing | Per-agent model field | L0 skill-router + adaptive overrides |
-| Decision logic | None (topic outlines) | Phase-aware, escalation chains, gates |
-| Context management | Native Claude Code only | context-engine, session-bridge, worktree |
-| Metrics | None | H3 metrics-collector, analytics |
-| Language coverage | 26 language agents | 0 (language-agnostic by design) |
-| Infrastructure depth | 16 agents | @rune/devops (6 skills) |
-| Business coverage | 11 agents | ba + marketing + @rune-pro/product |
-| Distribution | Plugin + installer | Plugin + 5-platform compiler |
-| Multi-platform | Claude Code only | Claude Code, Cursor, Windsurf, Antigravity, Generic |
-| Quality per agent | Shallow (checklists) | Deep (decision trees, escalations) |
-
----
-
-## Gaps to Fill in Rune
-
-### High Value (should adopt)
-
-| What | From VoltAgent | Rune Action |
-|------|---------------|-------------|
-| Tool permission field | `tools:` in frontmatter | Add to SKILL.md format |
-| Model inherit option | `model: inherit` | Add to skill-router |
-| Agent catalog commands | /subagent-catalog:search/list | `/rune agents list/search` |
-| Chaos/resilience testing | chaos-engineer | Add to @rune/devops |
-| Quant finance | quant-analyst (opus) | Add to @rune/trading |
-| Compliance auditing | compliance-auditor (opus) | @rune-pro/legal pack |
-
-### Medium Value (consider)
-
-| What | From VoltAgent | Rune Action |
-|------|---------------|-------------|
-| LLM architecture | llm-architect, prompt-engineer | Enrich @rune/ai-ml |
-| Legal/contracts | legal-advisor | @rune-pro/legal pack |
-| DX optimization | dx-optimizer | Could enhance onboard or audit |
-| Self-installer | agent-installer meta-agent | Nice DX for community packs |
-
-### Skip (not worth copying)
-
-| What | Why Skip |
-|------|----------|
-| 26 language specialists | Claude knows languages natively — marginal value, major bloat |
-| Duplicate agents | Quality issue — ml-engineer vs machine-learning-engineer |
-| Shallow checklist format | Rune's gate-driven approach is fundamentally better |
-| Fake progress metrics | Misleading, adds no real value |
-
----
-
-## Subagent Benefits (confirmed by analysis)
-
-1. **Context isolation** — each agent runs in own context window, main agent stays small
-2. **Tool permissions** — read-only agents (scanner, reviewer) can't accidentally write files
-3. **Model routing** — haiku for search ($0.001), sonnet for code ($0.01), opus for architecture ($0.05)
-4. **Specialization** — focused instructions = better output than "do everything" agents
-5. **Parallel execution** — independent agents can run simultaneously
-6. **Token efficiency** — load 1 agent file (~50 lines) vs entire skill mesh (~15,000 lines)