# Portfolio Readiness — To-Do Notes
## Completed
- [x] Author a concise case study (`docs/portfolio/case-study.md`): task, baseline, frameworks/gates applied, before/after quality or latency metrics, screenshots/logs, and any usage counts (installs, internal users).
- **Done**: Created `docs/portfolio/case-study.md` with recursive prompt engineering narrative
- **Content**: Problem/Solution/Results structure, 10+ quantified metrics, demo commands, technical trade-offs
- **Dogfooding**: Case study refined using `>>readme_improver` prompt with inline gates
- [x] Surface engagement signals despite low discussions: npm downloads, install counts, or internal usage stats; add to the case study and brief mention in `README.md`.
- **Done**: Added GitHub stars badge + collapsible Ecosystem section linking Smithery/Playbooks
- **Current metrics** (Dec 2025): 115 stars, 27 forks, 228 npm downloads/month
- **Indexed on**: [Smithery](https://smithery.ai/server/@minipuft/claude-prompts-mcp), [Playbooks](https://playbooks.com/mcp/minipuft-claude-prompts), DeepWiki
- **Case study framing**: "115 GitHub stars, 27 forks in specialized MCP ecosystem; indexed on major directories"
- [x] Create demo chains for video showcases
- **Tech Evaluation Chain** (`>>tech_evaluation_chain`) - 4-step library/framework evaluation for developers
- **Research Chain** (`>>research_chain`) - 4-step general research workflow for non-developers
---
## Demo Videos
### Video Strategy
| Client | Priority | Target Audience | Focus |
|--------|----------|-----------------|-------|
| **Claude Code** | Primary | Developers, Twitter trend followers | Full features: hooks, chains, gates, hot-reload |
| **Claude Desktop** | Secondary | Non-developers, enterprise | Clean MCP basics, prompt management |
| **Cursor** | Optional | IDE-focused devs | 1-click install, inline workflow |
**Skip Gemini CLI** - Hooks are Claude Code-specific, not portable.
---
### Video 1: Claude Code Demo (5-7 min)
**Target**: Developers, trending Claude Code audience
**Script Outline**:
1. **Opening (30s)**
- "I built an MCP server that turns prompts into a programmable system—versioned, hot-reloadable, with quality gates."
2. **Quick Install (30s)**
```bash
# First time: add the marketplace
/plugin marketplace add minipuft/minipuft-plugins
# Install the plugin
/plugin install claude-prompts@minipuft
```
- Show hook catching `>>syntax` if typed incorrectly
3. **Chain Demo: Tech Evaluation (2 min)**
```
>>tech_evaluation_chain library:'zod' context:'API validation in Express'
```
- Step 1: Library Overview (quick scan)
- Step 2: Technical Deep Dive (BLOCKING gate - must verify claims)
- Step 3: Integration Assessment
- Step 4: Recommendation with code examples
- **Key moment**: Show gate blocking on unverified claims
4. **Inline Gates (1 min)**
```
>>readme_improver content:'...' :: 'under 500 words' :: 'include Quick Start'
```
5. **Framework Injection (30s)**
```
@CAGEERF >>deep_research topic:'MCP protocol internals'
```
6. **Meta-Prompt: Create Your Own (1 min)**
```
>>create_prompt name:'my_validator' purpose:'validate API responses'
```
- Show wizard flow → validation → auto-creation
7. **Hot Reload (30s)**
- Edit prompt YAML in editor
- Run immediately without restart
8. **Closing (30s)**
- "Prompts as code. Version controlled. Quality gated. Hot reloadable."
---
### Video 2: Claude Desktop Demo (3-4 min)
**Target**: Non-developers, enterprise users, general audience
**Script Outline**:
1. **Opening (20s)**
- "Turn Claude into a structured research assistant with reusable prompts."
2. **Setup (30s)**
- Show JSON config in Claude Desktop settings
- Quick server startup
3. **Research Chain Demo (2 min)**
```
>>research_chain topic:'best practices for remote team management' purpose:'policy update'
```
- Step 1: Initial Scan (map the landscape)
- Step 2: Deep Investigation (BLOCKING - must cite sources)
- Step 3: Synthesis (reconcile findings)
- Step 4: Action Plan (concrete recommendations)
4. **Prompt Library (30s)**
```
resource_manager(resource_type:"prompt", action:"list")
```
- Show available prompts by category
5. **Closing (20s)**
- "Your prompt library, managed and versioned."
---
## Demo Chains Created
### Tech Evaluation Chain (Claude Code)
**Location**: `server/resources/prompts/development/`
| Step | Prompt | Gate | Purpose |
|------|--------|------|---------|
| 1 | `library_overview` | (advisory) | Quick health check |
| 2 | `technical_deep_dive` | **Verified Claims** (BLOCKING) | Rigorous analysis |
| 3 | `integration_assessment` | (advisory) | Compatibility check |
| 4 | `tech_recommendation` | **Actionable Output** | Go/no-go decision |
**Invoke**: `>>tech_evaluation_chain library:'zod' context:'API validation'`
### Research Chain (Claude Desktop)
**Location**: `server/resources/prompts/analysis/`
| Step | Prompt | Gate | Purpose |
|------|--------|------|---------|
| 1 | `initial_scan` | (advisory) | Map knowledge landscape |
| 2 | `deep_investigation` | **Source Citations** (BLOCKING) | Evidence gathering |
| 3 | `research_synthesis` | (advisory) | Reconcile findings |
| 4 | `action_plan` | **Actionable Recommendations** | Concrete next steps |
**Invoke**: `>>research_chain topic:'remote work' purpose:'policy update'`
---
## Pending Tasks
- [ ] Add demo video links near Quick Start in `README.md` and to résumé/LinkedIn notes (keep URLs stable)
- [ ] Record Claude Code demo video (5-7 min)
- [ ] Record Claude Desktop demo video (3-4 min)
- [ ] Test both chains end-to-end before recording
- [ ] Create thumbnail/cover images for videos
- [ ] Add video timestamps to descriptions
- [ ] Cross-post to relevant communities (MCP Discord, Claude Code discussions)
---
## Key Selling Points to Emphasize
1. **Chains** - Multi-step workflows with state passing
2. **Gates** - Blocking vs advisory quality validation
3. **Frameworks** - Methodology injection (@CAGEERF, @ReACT)
4. **Hot Reload** - Edit YAML, run immediately
5. **Meta-Prompts** - `>>create_prompt` for self-improvement
6. **Version Control** - Prompts as code in git
7. **Hooks** (Claude Code only) - Catches syntax errors, reminds about chain state
