SkydiveAI coworkers that actually get work done
Promoted
BotBugger
Two AI engines hunt bugs — and cross-check each other
9 followers
Two AI engines hunt bugs — and cross-check each other
9 followers
Claude Code and Codex analyze your code adversarially and verify each other's findings. Real bugs, concrete fixes.
one thing that would make this way more useful for me would be the ability to whitelist certain false positive patterns per project, since style debates between the two models will probably come up a lot in larger codebases.
@cosgun_hak89117 Spot on, Hakan — the style debates are the real noise in bigger repos, and muting those per project is exactly what Botbugger should do. I'm going to build a per-project "known/accepted" list so flagged findings stop resurfacing. Shipping a first version in the next few days — I'll ping you here when it's live. Thanks for thinking it through! 🐛
@cosgun_hak89117 t's live 🔕 Open a finding → "Mute this kind" → keyword. Matching findings are dismissed automatically, retroactively and on future runs; delete the pattern to bring them back.
Fun part: I ran the feature through Botbugger's own cross-check, and the second engine caught a real trap — plain substring matching meant muting const would also have silenced "SQL injection through unsafe query con-struction". So it's whole-word matching now, and critical/high are never auto-muted.
Let me know if it holds up on a big repo. Best feedback of the launch 🙏
ran it on a gnarly legacy module and the two agents actually caught a race condition i had been chasing for days, with a fix that just worked. feels like a second pair of senior eyes without the meeting.
Finally got two AI agents to argue with each other on my behalf and they actually found a race condition I had missed for weeks. The mutual verification step feels like having a second pair of eyes that refuses to sign off until the fix really makes sense.