Is SpecKit *really* maintained?

[massimogentilini]

With the new year coming and Den that left Microsoft to join Anthropic I'm a little worried about how the project is (NOT) progressing.

In the last month we had 22 PRs opened and exactly zero commits and the more time passes the more difficult it would be to merge everything on a new version.

Can someone from GH or Microsoft take care of managing this code base or comment if SpecKit as a tool is being abandoned in favour of adding it as a first class feature to VSCode or GitHub Copilot (totally reasonable from my point of view)?

Has some of you connection with GH or MS teams that can get in touch and elicit their point of view?

I feel that SpecKit and the way of working it impose on a project are a real game changer for any organization wanting to leverage coding agents in a proper manner, but if SpecKit is not maintained properly than I need to move on some different tooling for SDD, I really did not liked BMAD but I can change idea, if needed.

From my point of view SpecKit as a first class citizen of Copilot would be good, but I can understand that there are a lot of people using different assistants and having something generic is a strong motivation to progress the tool, but we need momentum to keep working successfully with SpecKit.

[crisweber2600]

@jflam do you have any thoughts about this?

[smith153]

There are newer tools coming up. I'm currently looking into https://github.com/obra/superpowers

[massimogentilini]

Seems to be that it does not support GitHub Copilot but other tools, correct?

[TabishB]

Hey all, I'm the maintainer of OpenSpec! It's a little sad to see this project wane, especially one that's open-source.

I was very keen to see where a project with a big backing like Microsoft could really take this space as I think with coding agents evolving, the usage of specs is becoming more and more important.

While we're not as big as Microsoft and we're not a 1:1 of spec-kit feature wise, we are relatively popular in the space at 18K+ stars and house a lot of extremely happy spec-kit users who have moved over.

We're also not sitting still, we're trying a lot of new things and we're constantly trying to figure out better experiences for users.

Check out some of the fun things we're trying here: https://github.com/Fission-AI/OpenSpec/blob/main/docs/opsx.md

Everything in the above doc is a direct result of months and months of community feedback.

While we can't promise everything will be the same as spec-kit, we think you might still have a fun time :)

[hillsmao]

The OpenSpec development framework is absolutely elegant! I’ve successfully built multiple projects with it. It’s document-driven and not complicated at all — perfect for non-technical product managers like me!

[atman-33]

I also use OpenSpec all the time. Thank you.

[BretJohnson]

The https://github.com/Fission-AI/OpenSpec/blob/main/docs/experimental-workflow.md page now shows a 404. Is there a new link that replaces it?

[indcoder]

I think that's because it's no more experimental https://github.com/Fission-AI/OpenSpec/blob/main/docs/workflows.md cc @TabishB to confirm, please

[TabishB]

@indcoder yes you're right! we've switched over, we've re-architected a lot of the internals to a new artifact based system that should allow us and our users to iterate and experiment with different approaches to spec-driven development!

@BretJohnson The equivalent to that links is now: https://github.com/Fission-AI/OpenSpec/blob/main/docs/opsx.md

[TyrelCB]

Spec Kit was dead to me after: 1) Not delivering a functional feature - nice spec though!! 2) Using all my premium tokens - 5x token usage over a vibe code that just worked 3) plan mode with claude code came out

[pasevin]

Congrats, you pointed out all the reasons for when not to use Spec Kit, now that doesn't mean it's not useful. I hope this project won't get abandoned, as it really resonates with my workflow and that of many others.

[mnriem]

Expect to see a flurry of activity soon

[massimogentilini]

Seeing your work I've quite inflated expectations.

[BretJohnson]

Are you working on a spec-kit fork? Just curious to learn more.

[morpheus9393]

I just want to start out with one of both. What are the pros and cons of Spec Kit and OpenSpec? And I read that OpenSpec is more for "brownfield". Does that mean it "does not make sense" or is much harder to implement for developing a new piece of software from 0 compared to e.g. Spec Kit?

Thanks for sharing your opinions!

[pasevin]

You should probably create a new discussion for this :) it would attract more people.

[massimogentilini]

you need to test both. OpenSpec for sure seems more "fluid" but to me is very good for teams where this "fluidity" is in the team as well, speckit seems to be more rigid but it's not something bad per se, keeping spec/clarify phase very well separated to the plan/task/implement from my point of view makes things easier for teams with a good separation of duties between analysts and developers.

And, at this precise moment, OpenSpec has a much more active community, for sure (things can change if MS and GitHub restart pushing SpecKit).

[howznguyen]

I've been dealing with the same frustration. Moved away from Spec Kit a while ago not because of the maintenance issue, but because I kept running into a deeper problem — AI forgets everything between sessions. Great specs, but every new chat starts from zero.

That's what pushed me to build something different. Instead of managing spec lifecycle, I focused on context persistence — AI reads project docs and task references automatically via MCP, no copy-pasting.

Still early stage and rough around the edges, but if anyone's exploring alternatives: https://github.com/knowns-dev/knowns
Curious what others here ended up switching to?

[morpheus9393]

There are many solutions out there for giving the AI a memory. From simple text (.md) based solutions, over Knowledge Graph systems to the premium solution formed by vector DB e.g. Qdrant. Vector databases also enable semantic searches, not just lexical searches.

[jbaruch]

For those looking for alternatives, I've been working on Intent Integrity Kit (IIKit) — a fork that extends Spec-Kit with cryptographic verification to prevent a critical failure mode: circular validation (where AI modifies tests to match buggy code).

What's different from Spec-Kit:

1. Skills instead of Prompts — IIKit migrates from prompt files to the skills format. Why does this matter?

   • Better alignment: Agents follow skills more reliably than raw prompts
   • Standardized: Skills are a recognized format across multiple AI coding assistants
   • Better distribution: Skills are distributed via Tessl, a proper package registry with versioning, dependencies, and 2000+ tiles for libraries/frameworks

2. Assertion Integrity — SHA256 hashing locks your Given/When/Then scenarios before implementation. If tests get modified, implementation is blocked until you go back to the spec. This is a new phase (/iikit-05-testify) that generates test specifications from requirements and locks them cryptographically.

3. Runtime Knowledge via Tessl MCP — During implementation, IIKit queries Tessl's library documentation in real-time via MCP. This means the AI uses current APIs and patterns (React 19, not React 16 from training data). The /iikit-03-plan phase auto-discovers and installs relevant tiles for your tech stack.

4. Cross-Platform — Full Bash + PowerShell parity, works with Claude Code, Codex, Gemini, OpenCode.

The core workflow builds on the phased approach Spec-Kit pioneered (constitution → specify → clarify → plan → tasks → implement), with added guardrails to ensure tests actually validate your requirements, not the AI's bugs.

Intent ──▶ Spec ──▶ Test ──▶ Code
       ↑       ↑        ↑
       │       │        └── hash verified (no tampering)
       │       └─────────── Given/When/Then locked
       └─────────────────── clarified until aligned

Links:

• GitHub: https://github.com/intent-integrity-chain/kit
• Install: tessl install tessl-labs/intent-integrity-kit
• Methodology: https://github.com/intent-integrity-chain/whitepaper
• Tessl registry: https://tessl.io

Happy to answer questions or hear feedback!