SPECTRE is an agentic coding workflow for people building products with Claude Code and other AI coding agents. It gives you a fixed sequence of steps, from Scope and Plan through Execute, Clean, Test, Rebase and Evaluate, so runs stay predictable across different features and codebases instead of turning into one off experiments every time. It is open source on GitHub and free to plug into your existing setup.
π₯ Our Take: I like that this just writes the rules down. No magic, no vibes, just the same routine every time an agent touches your code. Scope it, plan it, run it, clean it up, test it, rebase, look at the result. If you are using agents regularly, that kind of boring structure is exactly what keeps things from getting weird.