satiricalguru

Fast Jev Agents — Development skill for Claude Code

Development community

High-performance verbatim context compaction for coding agents (Claude, Codex, Antigravity, Gemini, OpenCode).

How to install Fast Jev Agents

This entry records only its repository, not the path inside it, so there is no exact command to give. Open satiricalguru/Fast-jev-agents and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Fast Jev Agents does

High-performance verbatim context compaction for coding agents (Claude, Codex, Antigravity, Gemini, OpenCode).

Alternatives in Development

Hapi — App for Claude Code / Codex / Gemini / OpenCode, vibe coding anytime, anywhere 3.6k ★
Agent Of Empires — Manage multiple Claude Code, OpenCode agents from either TUI or Web for easy access on mobile 3.1k ★
Hive — Browser-native hive-mind for CLI coding agents — Claude Code, Codex, Gemini, and OpenCode collaborate as real 473 ★

README

Fast-jev-agents

**Continuous, Verbatim Context Compaction for Autonomous Coding Agents**

*Never lose an exact line number, compiler error, or user constraint to lossy LLM summarization.*

[![npm version](https://img.shields.io/badge/npm-v0.3.0-blue.svg?style=flat-square)](https://www.npmjs.com) [![TypeScript](https://img.shields.io/badge/TypeScript-5.7+-3178C6?style=flat-square&logo=typescript&logoColor=white)](https://www.typescriptlang.org) [![Tests Passing](https://img.shields.io/badge/Tests-50%2F50%20passing-brightgreen?style=flat-square)](https://github.com/satiricalguru/Fast-jev-agents) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg?style=flat-square)](https://opensource.org/licenses/MIT) [![Supported Agents](https://img.shields.io/badge/Agents-Claude%20%7C%20Codex%20%7C%20Antigravity%20%7C%20Gemini%20%7C%20OpenCode-blueviolet?style=flat-square)](#supported-coding-agents)

Why Verbatim? • Optimizations • Quickstart • Agent Integrations • CLI Tool • Configuration • Contributors

The Problem: Lossy Summarization Breaks Agents

When an AI coding agent runs for 20+ turns, its conversation context approaches LLM window limits. Standard agent frameworks solve this with **summary compaction**: asking an auxiliary model to write a prose summary of older turns.

[!WARNING] **Summary Compaction is Destructive**:

File paths (src/core/auth/tokens.ts becomes "the auth module")
Exact error traces (Expected 200 OK, got 403 Forbidden at line 48 vanishes)
Strict user constraints ("Never edit files under src/generated") are often dropped or hallucinated away
Re-running tasks becomes erro