performance-audit
π―Skillfrom namesreallyblank/clorch
Analyzes code performance metrics, identifies bottlenecks, and suggests targeted optimizations for improving computational efficiency and resource utilization.
Installation
npx skills add https://github.com/namesreallyblank/clorch --skill performance-auditSkill Details
Overview
# Clorch
[](https://github.com/namesreallyblank/Clorch/releases)
[](LICENSE)
Claude + Orchestration β A cognitive architecture for Claude Code with RLM-native processing
```
βββββββββββββββββββββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββββββ
βββββββββββββββββββββββββββββββββββ
Skills β Agents β Hooks β Memory β RLM β Ralph
Parallel Orchestration
βββββββββββββββββββββββββββββββββββ
```
What is Clorch?
Clorch is an orchestration layer for [Claude Code](https://docs.anthropic.com/en/docs/claude-code) that transforms it from a single-threaded assistant into a multi-agent system with persistent memory, specialized skills, and intelligent context management.
Think of it as an operating system for AI-assisted development:
- Without Clorch: You interact with one Claude instance that forgets everything between sessions, handles all tasks generically, and struggles with large codebases.
- With Clorch: An orchestrator analyzes your request, spawns specialized agents in parallel, loads relevant expertise, remembers past decisions, and manages context so it never overflows.
Clorch is built on the RLM (Recursive Language Model) paradigm from [arXiv:2512.24601](https://arxiv.org/abs/2512.24601)βtreating large contexts as data to be processed recursively rather than crammed into a single prompt. This enables Claude to handle codebases and research tasks that would otherwise exceed its context window.
Who is it for?
- Developers who use Claude Code daily and want more power
- Teams managing large codebases where context overflow is a constant problem
- Anyone who wants Claude to remember decisions, follow conventions, and work in parallel
---
Two Modes of Operation
Clorch provides two complementary modes for different workflows:
| Mode | Use Case | How It Works |
|------|----------|--------------|
| Interactive | Active development, debugging, exploration | Orchestrator spawns parallel agents, manages context, returns synthesized results |
| Ralph (Autonomous) | Large tasks, AFK coding, overnight batch work | External bash loop runs fresh Claude per iteration until all tasks complete |
```
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β INTERACTIVE MODE β
β You: "Fix the auth bug" β
β β β
β Clorch: scout β investigate β sleuth β root cause β spark β fix β
β β β
β You: Review and approve changes β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β RALPH MODE (Autonomous) β
β You: /ralph "implement user dashboard with tests" β
β β β
β Clorch: Analyzes codebase, creates task list, opens terminal β
β β β
β Ralph: Fresh Claude per iteration β completes all tasks β
β β β
β You: /ralph validate β review results β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
```
Ralph is named after Ralph Wiggu
More from this repository6
Automates web browser testing and interaction workflows using Playwright, enabling Claude to programmatically control browser sessions and perform complex web-based tasks.
Manages project-specific lifecycle events and triggers custom actions during different stages of project development or interaction.
Coordinates and manages parallel AI agents, dynamically spawning specialized skills to handle complex development tasks while maintaining persistent context and memory across interactions.
Closes or terminates the current Claude Code session or interaction within the Clorch orchestration framework.
Enhances frontend design by generating aesthetically pleasing color palettes, typography recommendations, and UI layout suggestions for web interfaces.
Recursively processes large language model contexts by breaking them down into manageable chunks and systematically analyzing or transforming complex textual data.