Yoshi — Git skill for Claude Code
Context-pruning proxy for Claude Code and Codex: Jev judges which history is still needed, measured not claimed.
How to install Yoshi
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open pedronauck/yoshi and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Yoshi does
Context-pruning proxy for Claude Code and Codex: Jev judges which history is still needed, measured not claimed. POC here now, heading soon into https://github.com/compozy/compozy
Alternatives in Git
- Perfup — Autonomous performance optimization: research, PoC, benchmark, implement, review, PR 3.6k ★
- Icpg Bootstrap — /icpg-bootstrap — Bootstrap from Git History 705 ★
- MiniCode Roadmap — MiniCode already has a usable lightweight terminal coding workflow, but there is still a visible gap between t 388 ★
README
Yoshi
**A local context-pruning proxy for Claude Code and Codex.**
Jev judges historical context in many small requests. Yoshi applies the validated omissions while keeping the protocol intact, then forwards the request to Anthropic or OpenAI and streams the response back unchanged.
[](https://bun.sh) [](tsconfig.json) [](#status-and-limits) [](docs/BENCHMARKS.md)
[!WARNING] **Yoshi is a proof of concept. The real version is heading into [CompozyOS](https://github.com/compozy/compozy).** CompozyOS is an open-source operating system for AI agents: it runs Claude Code, Codex and other ACP agents as durable background sessions. There, context pruning becomes automatic for every session, the session goal and plan feed the judge, and judging runs between turns instead of on the request path, which removes the latency you will see below. If that is what you came for, [star CompozyOS](https://github.com/compozy/compozy) and follow along.
What it does
Coding agents resend their whole conversation on every request. Most of that history is completed tool input, tool output and teammate reports that the model no longer needs. Yoshi sits between the client and the provider and:
- Finds removable spans in history: completed
Bash,WriteandEditarguments, tool results, Claude teammate reports and old exchanges. The current instruction, recent turns, errors and opaque blocks (thinking, signatures, media) are always kept. - Asks Jev, one span per request, with up to eight requests in flight. Each request carries the current task, earlier user cons
Related Skills
Copilot Proxy API
Self-hosted GitHub Copilot bridge for Claude Code and Codex CLI with OpenAI Responses, Anthropic Messages, str
Advanced Code Review Context
Advanced Code Review Phase 2: Context Analysis - load previous reviews, PR history, declined items
Restart Sop
Run the Agent SOP session start checklist. Reads all context files, checks git history, flags inconsistencies,
Fashion Coach
Personal stylist for men, as a Claude Code skill. Install: git clone into ~/.claude/skills/fashion-coach, then
Koto
Move conversations between Claude Code, Codex, Cursor, and more. Keep your conversation history, branch into d
Token History
Token usage is the next contribution graph — snapshot yours to git daily, show it anywhere as self-updating SV
Related Agents
Patch Engineer
Drafts a minimal idiomatic unified diff that closes exactly the cited bound, then verifies it by re-running th
Dependency Drift
Verifies that code — generated by @zanix/cli, or cited/claimed by a skill's own examples — still compiles agai
Swe Dev Change Assurance Engineer
Reviews consequential changes by asking which prior claims and evidence still apply, what became stale or cont