Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Notes from the SF Safety Scene(12gramsofcarbon.com)
    discuss
  2. The Economics of Open-Weight Inference(ornn.com)
    discuss
  3. Texas police department ordered to close for failing to provide public benefit(dallasnews.com)
    discuss
  4. The draft AI code of conduct forbids me from saying 'I don't know'(ilands.ai)
    discuss
  5. Time to Spend Tokens or Meditate?(inmve.github.io)
    discuss
  6. It's the Fun(scottsumner.substack.com)
    discuss
  7. SEO Content Brief: What to Include for Better Rankings(briefiq.io)
    discuss
  8. Expat 2.8.5 released, fixes vulnerability CVE-2026-93990(hartwork.org)
    discuss
  9. A First Futamura Projection(veitheller.de)
    discuss
  10. Show HN: IntelliChat minimalist, open-source UI for local and cloud AI(github.com/intelligentnode)
    discuss
  11. Differential Equations, an Interactive Introduction(chapterpal.com)
    discuss
  12. A Simple Guide to Calm UI(maxschmitt.me)
    1comments
  13. Anthropic at $2T isn't far-fetched(ft.com)
    1comments
  14. Firedrill: Stateful tool simulation for AI agents(github.com/firedrill-tools)
    discuss
  15. Carefully Applied: Resume and LinkedIn Rewriting(carefullyapplied.com)
    discuss
  16. iPhone 15 Pro and 16 Settlement for Apple Intelligence taking claims – US ONLY(smartphoneaisettlement.com)
    discuss
  17. Nvidia Isaac ROS 5.0: agentic, open-source robotics development(nvidia.com)
    discuss
  18. Toadstools and Toxins(aeon.co)
    discuss
  19. Remote Code Execution (RCE) in a DoD Website(hackerone.com)
    1comments
  20. People Training OpenAI's AI Fired for Using AI to Train the AI(404media.co)
    discuss
  21. Worker Previews: isolated preview for every change your agent makes(cloudflare.com)
    discuss
  22. Websites that read in the terminal: the TermWeb standard(andros.dev)
    discuss
  23. Devin AI and SWE-2 First Impressions(catalins.tech)
    discuss
  24. George Lucas Museum Review: A Bold Throwback to Gilded Age Patronage(hollywoodreporter.com)
    discuss
  25. AI Is Antithetical to Learning(jola.dev)
    discuss
  26. Too many agents, one Mac(mcclowes.com)
    1comments
  27. Muse AI(cnbc.com)
    discuss
  28. Show HN: EnvSeal-CLI – Git for Secrets, Offline Git-Native Secret Manager(github.com/viswajith275)
    2comments
  29. Batteries that safely break down in GI tract could improve ingestible devices(news.mit.edu)
    discuss
  30. Vortex: One Format for Any Shape(spiraldb.com)
    discuss

Show HN: Nimblegate – Git push guardrails for AI agents

4 pointsby 1h agogithub.com
1 comments
1h agoHN ↗

I built nimblegate to check code faster, own and AI agent generated. Let agents generate your code and check every diffs for basic mistakes like test credentials pushed in code. This is exhausting time consuming, especially after long coding session. Nimblegate sits between your dev machine and final repo, self-hosted one-container docker(quick setup). Set your rules what goes through and re-bounced back to fix without manual checking and focus real problems. Auto-pr bounch catched mistakes looped automatically back to user or AI agent (multi agent/swarms) to fix.

Nimblegate catch mistakes exactly how you defined rules every time, AI agents has no access server files only git-shell what check and relay code into your final repo. Traffic protected ssh key and PAT (personal access token) dev machine (ssh) -> gateway and only gateway has write access(PAT) -> Github/Gitea etc. where is your final repo.

There is two modes what admin can set:

1. observe all goes through, shows dashboard what would blocked

2. gating mode what blocks based on rules.

License: PolyForm Noncommercial, free for personal and non-profit use, $10/month or $99/year per company for commercial use.