oh-my-taiyiforge

Why we included this project

TaiyiForge fits teams and solo developers who are tired of letting coding agents improvise through a change. It installs a small engine and a shared skill set across Claude Code, Codex, Cursor, and OpenCode, so the same slash commands behave identically in each tool, and it tracks each change through a nine-stage pipeline from requirement and design to review and integration. Human approval gates are wired in at the points where someone should still sign off, which matters when nobody trusts the model to approve its own design or review. It also handles the context problem by managing tokens, enforces acceptance criteria, and can split a large requirement into modules and generate a full-stack skeleton when you switch to auto mode. If the pain is inconsistent behavior across AI tools or agent work that runs without oversight, this gives you a repeatable, auditable workflow instead of hoping the model behaves.