Skip to content

Latest commit

 

History

9 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 

Repository files navigation

JeredBlu Marketplace

Custom Claude Code resources from JeredBlu's YouTube channel.

Overview

This marketplace provides skills that work across multiple AI coding tools:

  • prd-creator: Create comprehensive PRDs through conversational discovery
  • agent-skill-evaluator: Security and safety evaluation for agent skills
  • mcp-evaluator: Security and privacy evaluation for MCP servers
  • frugal-fable: Keep a premium model (Fable/Opus) on the judgment and delegate token-heavy work to cheaper models (Haiku/Sonnet) — for research, building, testing, and debugging. Have Codex CLI too? It can offload slices to GPT-5.6 (Sol/Terra/Luna) as a second quota pool — consent-gated, and zero Claude usage for the offloaded work

Most skills utilize MCP servers for enhanced functionality and follow the open Agent Skills standard. frugal-fable additionally needs Claude Code's workflow + sub-agent orchestration for its full feature set (see the platform note under Usage).

Installation Options

Option 1: Claude Code Plugin Marketplace (Recommended)

1. Add Marketplace

/plugin marketplace add JeredBlu/jeredblu-marketplace

2. Install Plugin

/plugin install jeredblu-tools@jeredblu-marketplace

Option 2: Manual Download (All Platforms)

Download skills individually for manual installation:

Skill Download Description
PRD Creator prd-creator.zip Conversational PRD generation
Agent Skill Evaluator agent-skill-evaluator.zip Security evaluation for skills
MCP Evaluator mcp-evaluator.zip Security evaluation for MCP servers
Frugal Fable frugal-fable.zip Premium-model orchestration + cheap-model delegation, incl. cross-CLI Codex offload (full features: Claude Code)

Platform-Specific Installation

Claude Desktop

  1. Download the desired .zip file
  2. Open Claude Desktop
  3. Go to Settings > Capabilities > Upload Skill
  4. Select the downloaded zip file
  5. Repeat for additional skills

Claude Code (Local)

  1. Extract the zip file
  2. Drop unzipped folder to ~/.claude/skills in your project directory or global Claude config
  3. Restart Claude Code

OpenAI Codex CLI

  1. Extract the zip file
  2. Move contents to ~/.codex/skills/<skill-name>/ or .codex/skills/<skill-name>/
  3. Restart Codex

OpenCode

  1. Extract the zip file
  2. Move contents to .opencode/skills/<skill-name>/ (project) or ~/.config/opencode/skills/<skill-name>/ (global)
  3. Restart OpenCode

Configuration

The skills work best with MCP servers. Both are optional but highly recommended for full functionality.

Recommended MCP Servers

GitHub MCP Server (Recommended)

Bright Data MCP Server (Recommended)

Context7 MCP Server (Optional)

Install and configure these MCP servers following their official installation instructions.


Usage

PRD Creator

Create comprehensive Product Requirements Documents through conversational discovery:

/prd-creator

Or naturally:

Help me create a PRD for my new app idea
I need to document requirements for a feature
Create product requirements for [project description]

The skill will:

  • Guide you through discovery questions (problem, users, features)
  • Research relevant technologies via MCP servers
  • Generate structured PRD with all sections
  • Support iterative refinement

Output includes:

  • Problem statement and goals
  • User personas and stories
  • Functional/non-functional requirements
  • Technical considerations
  • Success metrics
  • Timeline and milestones

Agent Skill Evaluator

Evaluate the security of agent skills from various sources:

Evaluate this skill: https://github.com/username/skill-repo
Is this skill safe? https://example.com/my-skill.zip
Security assessment for this skill please: [attach .skill file]

The evaluator will:

  • Download and extract the skill
  • Analyze SKILL.md for prompt injections
  • Review scripts for malicious code
  • Search community feedback
  • Generate comprehensive security report with risk scoring

Features:

  • Prompt injection detection
  • Malicious code pattern matching
  • Hidden instruction scanning
  • Data exfiltration detection
  • Community validation
  • Risk scoring (0-100 scale)
  • Actionable recommendations

MCP Server Evaluator

Evaluate the security of MCP servers:

Evaluate this MCP server: https://github.com/username/mcp-server
Is this MCP safe to use? https://github.com/org/mcp-repo

The evaluator will:

  • Analyze repository metadata and activity
  • Review code for security vulnerabilities
  • Search for alternatives and comparisons
  • Gather community feedback (including Reddit with Pro Mode)
  • Generate detailed assessment with recommendations

Features:

  • Security vulnerability analysis
  • Privacy risk assessment
  • Code quality review
  • Alternative server discovery
  • Community feedback research (Reddit, forums, GitHub)
  • Multi-dimensional scoring
  • Usability assessment

Frugal Fable

Platform note: Full functionality requires Claude Code (a Fable-class runtime with workflow + sub-agent orchestration). On Claude Desktop, Codex, or OpenCode the written delegation guidance still applies, but the bundled budget-capped research workflow won't run.

Keep your most capable model on the thinking and push token-heavy volume to cheaper models:

/frugal-fable

Or naturally:

Be efficient with Fable on this — don't burn tokens
Research X but delegate the searching to cheaper agents
Build this feature but keep Fable on the architecture and review

The skill will:

  • Keep the session model (Fable/Opus) on decomposition, architecture, synthesis, and review
  • Delegate scans, fetches, bounded patches, and tests to Haiku/Sonnet with a conservative quality floor
  • Use a file-based "context firewall" so sub-agent output doesn't bloat the orchestrator's context
  • Reach for the bundled budget-capped frugal-research.js workflow for cost-bounded deep research

Lanes: research · build/dev · testing · debugging. As with any delegation, the orchestrator reviews delegated work before relying on it (the built-in quality gate) — so you stay in control of what ships.

Note: "Fable" is just the marquee example — the skill uses whatever model your session starts on as the orchestrator (Opus works identically) and delegates down from there.

Cross-CLI: Codex as a second pool (new in 1.2.0)

If you also have OpenAI's Codex CLI installed and logged in, frugal-fable can delegate slices to Codex models — GPT-5.6 Sol / Terra / Luna, tier-mapped onto the same routing rules (Sol ≈ Opus-class, Terra ≈ Sonnet, Luna ≈ Haiku). Why this is more than a nice-to-have:

  • It's a separate quota. Codex runs on your ChatGPT subscription, so every slice routed there costs zero Claude usage — the frugal move isn't just "cheaper model," it's "different bill."
  • Consent-gated. The skill never routes to Codex on its own: it detects the CLI, asks you once per session, and stays silent on machines without Codex. If you're away when it would ask, it proceeds on Claude models and mentions the offer in its summary.
  • Battle-tested mechanics. Delegation uses synchronous codex exec with file-based reports (no context bloat), verified recipes for build slices, adversarial second opinions, repo/PR reviews, and same-session amendments — including fixes for real gotchas we hit (resumed Codex sessions silently revert to your config defaults; the recipes re-pin the sandbox every time).

Try it naturally:

Delegate the heavy parts of this refactor to Codex
Get a second opinion from Codex on this diff
Build this, keep Claude on architecture, offload grunt work to my Codex quota

Requires: Claude Code as the orchestrator + codex CLI on PATH. Without Codex, nothing changes — the skill behaves exactly as before.


Graceful Degradation

Skills work without MCP servers but with reduced functionality:

Scenario Behavior
No GitHub MCP Uses web scraping for repository access
No Bright Data Uses built-in web search (limited)
No Pro Mode No Reddit scraping, basic search only
No Context7 Manual documentation lookup

Example: Without MCPs

User: "Evaluate this MCP: https://github.com/example/server"
Claude: Uses basic web scraping, can't access private repos,
        limited Reddit data, slower analysis

Example: With MCPs

User: "Evaluate this MCP: https://github.com/example/server"
Claude: Direct repo access, full code review, Reddit community
        feedback, comprehensive security scan

Requirements

  • Claude Code, Claude Desktop, VS Code, Cursor, Codex, or OpenCode
  • GitHub Personal Access Token (recommended)
  • Bright Data API token (recommended, for Reddit scraping)

Coming Soon

  • Custom slash commands
  • Agents
  • Additional skills

About

Custom Claude Code resources: skills, commands, and tools from JeredBlu's YouTube channel

Resources

Stars

61 stars

Watchers

2 watching

Forks

Releases

Packages

Contributors

Languages