duoduoler-ops/Table-skills

What it solves

It addresses the problem of losing critical context—such as finalized decisions, rejected options, and actual progress—when switching conversations during long AI tasks. It prevents the need to manually copy and paste entire chat histories to maintain continuity.

How it works

The project provides "Skills" (instructional sets and scripts) that can be installed into AI agents like Codex (Windows) or Claude Code. The primary tool, project-handoff, creates a structured handoff process where the agent saves goals, key files, and next steps into a PROJECT_STATE.md file. When a new conversation starts, the agent reads this state file to resume work accurately.

Who it’s for

Users of AI agents (specifically Codex on Windows and Claude Code) who manage complex, long-term projects that span multiple chat sessions.

Highlights

• Structured State Management: Saves only the essential progress and decisions rather than raw chat logs.
• Automated Reminders: Includes logic to suggest a handoff after a certain number of context compressions or when a significant project phase changes.
• Cross-Agent Support: Provides adaptation for both Codex and Claude Code via specific hooks and scripts.
• Customizable Hooks: Uses system-specific commands (PowerShell for Windows) to track state and trigger reminders.