Harvil1/codeAgent
This repo pitches CodeAgent as a Windows-first, handwritten Chinese clone of Claude Code with a full agent stack exposed for study and modification.
It describes a terminal AI coding assistant that can read code, edit files, run commands, search, and delegate work to sub-agents. The project claims built-in approval flows for destructive commands and out-of-whitelist writes, plus secret scanning, SSRF protection, injection detection, and optional Linux `bwrap` sandboxing. It also includes memory, skill learning from conversation traces, workflow automation, cron tasks, hooks, checkpoints, and multi-model support. Source: GitHub · LLM repos' note.
It describes a terminal AI coding assistant that can read code, edit files, run commands, search, and delegate work to sub-agents. The project claims built-in approval flows for destructive commands and out-of-whitelist writes, plus secret scanning, SSRF protection, injection detection, and optional Linux `bwrap` sandboxing. It also includes memory, skill learning from conversation traces, workflow automation, cron tasks, hooks, checkpoints, and multi-model support. Source: GitHub · LLM repos' note.
score 4