Everything Claude Codex: 11 Lessons From 11 Months in Claude Code

Описание и применение

Система Everything Claude Codex автоматизирует рутинные задачи и обеспечивает контекстуальное взаимодействие двух ИИ‑агентов, устраняя необходимость ежедневного вводного брифинга и выявляя скрытые сбои.

Задача

The author needed an assistant that remembered context across sessions and could automate routine tasks without manual briefing each day.

Результат

The system reduced daily briefing overhead, automated routine tasks, surfaced silent failures, and enabled seamless collaboration between two AI agents, improving productivity for a busy student.

Рабочий процесс

• Create and maintain hot.md with current context
• Implement session_context hook to load hot.md and recent summaries
• Define skills (/opps, /boot, /outreach, /ship, /video-notes) that read hot.md
• Add hooks for rules (e.g., no em dashes) and guards (dangerous commands)
• Set up doctor with checks for hooks, indexes, jobs, syncs
• Implement mcp_health.py to verify tool availability before calls
• Create /ship skill to commit, push, and report proof
• Synchronize two agents via agent-sync and AGENTS.md
• Sync notes to phone via phone-vault and GitHub Actions
• Configure vault-search for semantic, relational, and image queries
• Limit phone sync to session summaries, not raw chats

Доказательства

Every session began the same way: me explaining who I am, what I'm building, what I already tried, and what's due this week.

Источник

The biggest jump in quality I ever got didn't come from a better model. It came from one file.

Источник

My nightly search index failed 120 nights in a row. No error, no alert.

Источник

After every /compact it gets refreshed.

Источник