#492 · Primary category: AI Agents & Automation

J-Space-Cognition-Suite-V3.6

agent-skills ai ai-agent ai-agents claude-code codex cognitive-enhancement deepseek deepseek-harness developer-tools dsh dsh-plugin global-workspace hermes-agent inference-time-control j-space opencode react tailwindcss

J-Space Cognition Suite V3.7 - AI cognitive-enhancement Skills based on Anthropic's J-space global workspace research. | 哔哩哔哩:Tiger380 (UID 3494375382321675) — https://space.bilibili.com/3494375382321675

Project last updated:08/23/26

GitHub Stars

3.0K

Forks

215

Contributors

2

License

Apache-2.0

Why we included this project

A coding or research agent on a long, multi-step job tends to decay partway through: constraints leak across files, the model circles back to a route it already failed, and the result it hands back may never have been verified. J-Space Cognition Suite exists to catch exactly those failure modes. It treats the model's working memory as a deliberately managed workspace, and it does so at inference time only: no model weights change, no fine-tuning is needed. Text sets the operating frame, modules route the computation, and an optional local controller preserves durable state between task seams, so a few active items stay in focus while longer reasoning chains live in a compact ledger and results only come back after an explicit verification step. Because the suite is model-agnostic and ships as a Skill, you can load it selectively on top of Claude Code, Codex, OpenCode, or a DeepSeek harness without restructuring your whole stack.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category