Claude Night Market

Claude Night Market

Overview

Claude Night Market is a comprehensive Claude Code plugin marketplace providing 16 plugins with 126 skills, 114 commands, and 41 agents for software engineering workflows. The plugins are organized in architectural layers (Foundation, Utility, Domain Specialists, Meta) covering git operations, code review, spec-driven development, issue management, multi-LLM delegation, TDD enforcement, and session management. The ecosystem adds approximately 14.8k characters to system prompts and includes cross-session state persistence via Claude Code Tasks (v2.1.16+).

Problem Addressed

Key Statistics

Key Features

Plugin Architecture Layers

Core Plugins

Governance and Quality

Installation Methods

# Plugin marketplace
/plugin marketplace add athola/claude-night-market

# Install specific plugins
/plugin install sanctum@claude-night-market
/plugin install pensive@claude-night-market
/plugin install spec-kit@claude-night-market

# npx (alternative)
npx skills add athola/claude-night-market
npx skills add athola/claude-night-market/sanctum

Notable Workflows

Technical Architecture

Directory Structure

plugins/
  <plugin-name>/
    skills/
      <skill-name>/
        SKILL.md          # Agent-facing instructions
    commands/
      <command-name>.md   # Slash command definitions
    agents/
      <agent-name>.md     # Agent configurations
    hooks/
      <hook-name>.md      # Behavioral hooks

Cross-Session State (Claude Code 2.1.16+)

• attune, spec-kit, sanctum integrate with native Claude Code Tasks system
• Task creation on-demand with persistence via CLAUDE_CODE_TASK_LIST_ID
• war-room-checkpoint enables embedded escalation at decision points
• Fallback to file-based state for versions prior to 2.1.16

LSP Integration (v2.0.74+)

• Symbol search in ~50ms (faster than text search)
• Requires ENABLE_LSP_TOOL: "1" in ~/.claude/settings.json
• Compatible with language servers like pyright

Prompt Context Management

• ~14.8k character system prompt budget (limit: 15k)
• Enforced by pre-commit hook
• Modular designs and progressive loading to stay within limits

Relevance to Claude Code Development

Direct Applications

1. Plugin Architecture Patterns: Layered architecture (Foundation, Utility, Domain, Meta) provides clear separation of concerns for plugin ecosystem design.

2. Hook Governance: PreToolUse hooks for TDD enforcement and rigorous reasoning demonstrate behavioral guardrails without code changes.

3. Multi-LLM Delegation: conjure plugin patterns for routing tasks to Gemini/Qwen while retaining oversight show hybrid AI workflow design.

4. Session Management: sanctum's checkpointing and resume strategies address context loss between conversations.

5. Expert Routing: war-room's Type 1/2 reversibility framework provides structured approach to decision escalation.

6. Code Review Taxonomy: pensive's domain-specific reviews (architecture, bugs, API, math, Rust, shell, Makefile) show comprehensive review coverage.

Patterns Worth Adopting

1. Layer-Based Organization: Foundation (core utilities), Utility (resource management), Domain (task-specific), Meta (authoring tools) provides clear mental model.

2. Quality Gate Enforcement: Halting execution on failing tests before allowing skill/command creation enforces quality.

3. Stability Metrics: pensive's usage frequency and failure rate tracking identifies unstable workflows.

4. Progressive Depth Levels: /cleanup command with configurable depth levels allows graduated thoroughness.

5. War Room Pattern: Multi-expert consultation with reversibility scoring for high-stakes decisions.

6. Slop Detection: scribe's AI slop detector identifies AI-generated content markers for quality control.

7. Memory Palace Technique: Spatial knowledge organization for skill execution memory and PR review context.

Integration Opportunities

1. Skill Import: Compatible plugin format allows cross-marketplace skill sharing.

2. Hook Patterns: imbue's TDD enforcement hooks could inform this repository's quality gates.

3. Review Domains: pensive's review taxonomy (15 specialized reviews) could expand coverage.

4. Delegation Framework: conjure's delegation-core skill provides patterns for external LLM integration.

5. Session Persistence: sanctum's cross-session state patterns address conversation continuity.

Comparison with This Repository

References

Research Method: Information gathered from GitHub repository README, GitHub API for repository metadata (stars, forks, license, dates), marketplace.json for plugin details, and capabilities reference for skill/command/agent counts. Statistics verified via direct API calls on 2026-01-31.

Freshness Tracking

Review Triggers:

• GitHub stars milestone (250, 500, 1K)
• Major version bump (2.x.x release)
• New plugin layer additions
• Cross-session state API changes
• War-room framework updates
• New LLM delegation targets added
• Significant skill additions (20+ new skills)