shareAI-lab/learn-claude-code

Visual summary of the agent harness architecture

Open

#355 geöffnet am 8. Juni 2026

Auf GitHub ansehen
 (0 Kommentare) (2 Reaktionen) (0 zugewiesene Personen)TypeScript (9.907 Forks)batch import
good first issue

Repository-Metriken

Stars
 (60.513 Stars)
PR-Merge-Metriken
 (Durchschn. Merge 10T) (17 gemergte PRs in 30 T)

Beschreibung

Hi shareAI-lab team,

I’ve been studying this repo and sketched a visual summary of how I understand the agent harness architecture.

The diagram connects several concepts from the course, including system prompt assembly, memory, auto-compaction, permissions/hooks, background tasks, cron scheduling, task systems, subagents, agent teams, and worktree isolation.

I’m sharing it in case it is useful for the docs or for other learners. Happy to revise it if any part does not match the intended architecture.

Thanks for creating this repo. It helped me think more clearly about agent harness engineering.

Contributor Guide