LLM Wiki

Appearance

Built 26/04/19 07:55commit 447639a

Everything Claude Code GitHub Repository ​

中文 | English

Summary ​

This source captures the concrete affaan-m/everything-claude-code repository rather than just the two companion Claude Code articles. It shows the distribution layer behind the ideas: cross-harness packaging, Codex-specific config and AGENTS guidance, layered rules, hook control surfaces, and reusable skills such as continuous learning and strategic compaction.

Source ​

Key Contributions ​

  • Grounds the shorthand and longform Claude Code advice in a real repository surface instead of only in article prose.
  • Shows that the same harness philosophy is packaged across Claude Code, Codex, Cursor, Kiro, OpenCode, Gemini, and other agent environments.
  • Demonstrates that Codex support is repo-native here: .codex/AGENTS.md, .codex/config.toml, .agents/skills/, and [agents.*] roles are first-class assets rather than afterthoughts.
  • Makes the control plane legible through concrete repo assets: layered rules, hook lifecycles, plugin manifests, MCP defaults, installers, and reusable skills like continuous-learning and strategic-compact.

Practical Implications ​

  • Once an agent workflow stabilizes, maintainers often need to ship it as a multi-surface repository product rather than a single-tool setup gist.
  • Cross-harness reuse is not free abstraction: it requires explicit per-surface packaging and compatibility maintenance.
  • Repo-local AGENTS/config/rules/hooks surfaces can preserve one operating philosophy while still adapting to each harness's native control model.