penguin-harness
🐧 Unified and Stable RSI Platform
- agent
- agentic-ai
- ai
- build-tool
- claude-code
- deepseek
- deepseek-harness
- desktop
- harness
- llm
- rsi
- self-evolving
- Stars
- 2,753
- Forks
- 294
- + today
- +7
- Created
- 3mo
Ranking data as of October 16, 2026 (UTC).
Star History
Today, hour by hour
Overview
🐧 Unified and Stable RSI Platform It ranks #37 on GitTiger, gaining +7 stars on October 16, 2026 (UTC).
The project is written in TypeScript and has 294 forks. It was created 3mo ago.
What it is
penguin-harness is a TypeScript project under the Apache-2.0 license, per the supplied metadata. Its description calls it a unified and stable RSI platform, meaning a framework for improving an agent harness automatically. Topics include agent, desktop, deepseek-harness and self-evolving.
Capabilities
The README preview lists these, in the project's own framing: - Several RSI algorithms (Default, OPRO, APE, ACE and AWM) provided as plugins that share built-in benchmarks, a scoreboard and snapshots. - Building an agent app from a plain-language description, including its page and run instructions. - A small toolset aimed at open models.
The preview also makes a cost comparison with Claude Code on the authors' own data-analysis suite. That is the authors' claim and was not verified here.
Getting started
No installation or usage steps were supplied. See the repository README.
git clone https://github.com/Prism-Shadow/penguin-harness.git
cd penguin-harness
# see README for setupFrequently Asked Questions
Which RSI algorithms are included?
The README preview names Default, OPRO, APE, ACE and AWM as plugins.
Is the cost claim independently verified?
No. The preview states it about the authors' own suite, and nothing in the supplied evidence verifies it.
What language is it written in?
The supplied metadata lists TypeScript.