Claude is too hesitant to try third party... too hesitant, and prefers untested homebrew solution over repo with 150k stars.
retained text
Title. I'm testing [https://github.com/github/spec-kit](https://github.com/github/spec-kit) , and ask Sonnet to evaluate the repo. It reads the README, and **compares it against my 1 day old attempt at spec driven development** 😄 It has declined many other frameworks, before: three.js, etc, and **EVERY TIME** it gives me a plausible looking reason for "we are already doing this". I finally called its bullshit and am reporting this here. It literally wants to use a home-brew solution of less than 48 hours old is "ahead on most of what matters", over a repo with 150k stars: Verbatim response **Verdict: spec-kit is not worth adopting as a tool. A few of its ideas are worth borrowing; our spec workflow is ahead on most of what matters.** This is read-only research and I changed no repository. \[two caveats not wroth mentioning it here. **What spec-kit is.** It is github/spec-kit, MIT licensed, 13.5k forks and about 150k stars. It is a Python CLI (`specify`) that installs slash commands into a coding agent. The flow is constitution, specify, plan, tasks, implement, then converge. Each feature gets [`spec.md`](http://spec.md), [`plan.md`](http://plan.md), [`research.md`](http://research.md), [`data-model.md`](http://data-model.md), `contracts/`, [`quickstart.md`](http://quickstart.md) and `tasks.md`.
document fetched 2026-11-18 00:44 · canonical https://reddit.com/r/ClaudeCode/comments/1xym5ig/claude_is_too_hesitant_to_try_third_party_too