Oranquelui

Astra Jev Harness — Development skill for Claude Code

Development community

Agent Skills for Codex Desktop and Claude Code, plus a Codex CLI harness.

How to install Astra Jev Harness

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

What Astra Jev Harness does

Agent Skills for Codex Desktop and Claude Code, plus a Codex CLI harness. Jev-assisted context selection. / Codex Desktop・Claude Code向けAgent SkillsとCodex CLI用ハーネス。Jevでコンテキストを選別。

Alternatives in Development

Petdex — A public gallery of animated pets for Codex, Claude Code, DeepSeek Harness, Hermes, OpenCode, Gemini CLI, and 4k ★
Directional Prompting — Outcome-first plus directional language 138 ★
SpecFusion — 在 DeepSeek Harness / Claude Code / Cursor / Codex / Gemini CLI 里直接搜索 20 个中国开放平台的 65,600+ 篇 API 文档；零配置，支持 Skill 66 ★

README

Astra + Jev: Codex and Claude Code Agent Skills

[日本語](README-ja.md) · [Version 0.5.0](VERSION) · [Tagged releases](https://github.com/Oranquelui/astra-jev-harness/releases) · [Changelog](CHANGELOG.md)

This repository provides the **[`astra-jev-coding` Codex Agent Skill](Codex%20Desktop/skills/astra-jev-coding/SKILL.md) for Codex Desktop** and a separate harness for Codex CLI. Install the Skill to use Jev for context selection while the current Desktop conversation implements the task; the CLI workflow runs Codex separately. A separate **[`claude-jev-coding` Skill](Claude%20Code/skills/claude-jev-coding/SKILL.md) for Claude Code** applies the same context selection to the current Claude Code session.

Narrow large repositories locally, let **Jev judge the bounded candidates**, then let **your coding model write the code**.

A local coding harness for Codex CLI, Codex Desktop and Claude Code. Give it a task such as “fix pagination without changing the public API”; it snapshots eligible files, locally narrows oversized repositories, asks Jev which candidates the coding model needs, preserves dependencies, and records what was kept and why.

**Goal:** reduce Astra token consumption and combined inference cost while preserving coding correctness and avoiding extra turnaround time. File selection is a means to that goal. The Claude Code Skill extends the same goal to Claude; its savings are not yet measured.

**Experimental.** One synthetic CLI coding task at **Astra Extra High (`xhigh`)** used **27.0% fewer Astra input tokens** and **27.2% less at equivalent Standard API rates**, including Jev. Both modes passed the same six checks. Elapsed time was 7.3% shorter in this pair, but 34.6% longer in a separate `medium` pair. Each is one trial per mode, not a general speedup or Desktop result. [Measurements and limits](docs/BENCHMARKS.md).