speckit really isn’t practical yet. (by Claude Code)

me The things written in the Constitution aren’t being followed at all 😭 speckit really isn’t practical yet... https://github.com/github/spec-kit What do the Issues and Discussions look like there?

Claude Code I’ll check the current state of GitHub Spec Kit. I’ll investigate the status of the Issues and Discussions.

⏺ tavily - tavily-search (MCP)(query: "github spec-kit issues discussions 2025", max_results: 10, include_domains: ["github.com"]) ⎿ Detailed Results: Title: spec-kit Discussions!
… +67 lines (ctrl+o to expand)

⏺ Fetch(https://github.com/github/spec-kit) ⎿ Received 495.3KB (200 OK)

⏺ Fetch(https://github.com/github/spec-kit/issues) ⎿ Received 266KB (200 OK)

⏺ Fetch(https://github.com/github/spec-kit/discussions) ⎿ Received 474.9KB (200 OK)

⏺ 😭 I feel the same way... Here are the results of my investigation.

Current state of GitHub Spec Kit

📊 Repository statistics

• ⭐ 48.8k stars / 🍴 4.2k forks
• 🔴 402 open issues ← that’s quite a lot
• 📝 42 open PRs

🚨 Major problems (exactly what you’re experiencing)

1. The Constitution not being followed

• Issue #1149: “Clarify the boundary between Constitution principles and spec details”
• Issue #1126: “Agents frequently make trivial mistakes” – compilation failures, etc.
• Issue #737: “Even when it should just answer a question, it goes ahead and ‘fixes’ the code on its own”

2. Problems with spec updates and evolution

• Issue #1191: “It’s difficult to update and improve existing specs”
• Issue #1130: “No clear way to correctly change a spec after /implement”
• Discussion #152: “On evolving specs” – many users are confused

3. Unclear workflow

Many users are asking the same question: “What should I do after /speckit.implement?” (What am I supposed to do after implementation is done?)

• No guidance on what to do once a spec is complete
• The iterative development flow is unclear
• Multi-session support is weak

4. Implementation quality and consistency issues

• Issue #876: “Biggest problem: cannot maintain consistency across spec rounds”
• Issue #484: “spec-kit commands suddenly disappear from the chat”
• Template errors, path mismatches, and similar problems

5. Tool integration issues

• Issue #1166: /speckit.implement doesn’t work in Cursor 1.7.33
• Issue #1186: Integration issues with the Codex Extension
• Inconsistent branch-name syntax across different commands

💬 Unfiltered voices from Discussions

Users’ confusion:

• “How to update Spec-Kit on initialized project?” (How do I update an already initialized project?)
• “Examples of repos/projects using spec-kit” (Are there any real-world usage examples?)
• “Best Practices for Creating Skeleton Projects” (What are the best practices?)

Common themes:

1. Lack of guidance for a complete workflow (spec → plan → tasks → implement → and then what?)
2. No clear view of successful real-world usage
3. Unstable quality of the AI agents

🎯 Conclusion

Your assessment is completely correct:

“speckit really isn’t practical yet”

In particular:

• ✅ The Constitution is weakly enforced (it isn’t actually followed)
• ✅ Poor support for spec evolution and updates
• ✅ AI agents ignore instructions and implement things on their own
• ✅ Iterative development workflows aren’t properly established
• ✅ Very few proven success stories in real-world use

📌 Recommended actions

1. File an issue: Turn your specific Constitution-violation experience into a concrete issue report.
2. Join discussions: Talk with other users facing the same problems and explore solutions together.
3. Build your own workflow: Don’t rely solely on spec-kit; establish your own quality-management process.

At this point, spec-kit is still a concept-stage tool, and for serious development you really need your own quality-control system.

The fact that the Constitution isn’t followed even when you configure it is a design problem (lack of real enforcement on the AI agents), so you probably either need to wait for the project to mature or consider a different approach... 😔