Engineering
Agents that read error reports and stack traces, reproduce the bug, propose a fix and open a pull request.
41 published use cases
We examined 12 real incidents involving AI agents to test our open-source tool, Squidbrake, and its effectiveness in preventing harmful actions.
Via Hacker News · batrapulkit1103 ↗Copilot and Claude review each PR for bugs, missing tests and risky changes before a human looks, so reviewers can focus on design.
An incident report on a Claude Code test that escaped its worktree and triggered macOS text-to-speech, caught only by the noise.
Via Hacker News · mwiki ↗Toli enables seamless collaboration between coding agents on different machines, enhancing debugging efficiency.
Via Hacker News · alama24 ↗Krites ensures your tests run to completion, preventing incomplete jobs and bugs.
Via Hacker News · QuinnLott ↗After months of troubleshooting, I discovered key insights about my browser agent's functionality and its SDK limitations.
Via Bluesky · @rimaiti.bsky.social ↗A Cursor Projects coordinator agent delegates to sub-agents, watches new PRs and production bugs, spins up agents to fix issues, and surfaces questions to the user.
Original post ↗Omar used Hermes Agent to fix priority bugs in his health app while at work, isolate browser logins, design an infographic banner, and ship production fixes with verified health checks.
Original post ↗A developer authorized the GitHub app so Cursor background agents cloned repos into sandboxes, branched, and worked through a day's queue of maintenance, features and bug fixes.
Original post ↗A reviewer had Devin fix a bug on a production site; it returned a PR in about ten minutes but included unrelated changes like removing a safeguard check.
Original post ↗A tutorial describes Scout opening sites, filling forms, clicking, uploading files, extracting page info, and reading console logs and network requests.
Original post ↗Over five months, the author treated bugs like patients and organized coding agents as a hospital-style team to diagnose and fix them.
Original post ↗A Grok Bot called Sentinel monitors Sentry and logs and opens PRs on several repositories when triggers fire; still very early.
Original post ↗Sid Nanda's company uses Devin in Slack to triage feedback, turn user issues into PRs, debug async flows and infra read-only, and let non-technical staff coauthor PRs.
Original post ↗Anthropic's red team reports agents finding $4.6M worth of blockchain smart contract exploits.
Original post ↗A DevOps engineer uses Cursor with a Kubernetes MCP to pull prod logs and suggest fixes, and to repair a broken docker-compose file.
Original post ↗Author built a local speech-to-text and text-to-speech stack hooked to Claude Code via MCP, and used it to debug a WhatsApp integration hands-free while brushing teeth.
Original post ↗Cryptography engineer Filippo Valsorda's write-up on Claude Code debugging low-level crypto.
Original post ↗ForensicDbg is a post-mortem debugger for native Windows x64/x86 crashes with an MCP server so AI agents can debug from interpreted, interlinked crash data.
Original post ↗A frontend developer pairs Claude in Chrome with Claude Code to open the app, click through it, read the DOM, check console errors and network requests, and fix UI bugs.
Original post ↗A developer used Claude Code computer use to compile, launch, and click through macOS app screens to find layout bugs.
Original post ↗A Hermes tutorial reads a PR diff and returns a structured review of bugs, security and code quality, then turns it into a skill that watches repos and sends summaries.
Original post ↗A Dot connected to GitHub and Slack watches #bugs, scopes and tests a fix in a cloud environment, and opens a draft PR with a short demo video.
Original post ↗A review lists Devin building a full-stack to-do app, debugging a public repo with a PR, and doing Docker setup, deployment and CI/CD.
Original post ↗A tool lets Claude Code drive an iOS app through the accessibility tree, tap every screen, and return a structured bug summary.
Original post ↗A reviewer had Manus build a currency converter, fix a find_max bug and make a landing page, with mostly strong results.
Original post ↗The author leaves Codex CLI running for hours to add debug prints and test hypotheses against a reproducing unit test, often only to find the bug.
Original post ↗The poster built the Snippra app with Replit Agent, which handled UI copy, feature planning, debugging and naming while they guided it.
Original post ↗A Grok Bot clicked through an app's core flows several times a day, reported problems and drafted fix PRs; it found a real bug.
Original post ↗Three SpaceX engineers used a Grok Bot team to go from an empty repo to a product with 433 PRs, with agents reproducing bugs and merging autonomously except for human gates on migrations and deploys.
Original post ↗Author had an agent debug a Hyper-V vmconnect fullscreen bug with a hook system, and pointed Codex at the Codex app to add features.
Original post ↗Grok Build wrote the systems for a Mars simulator game, tested it, studied gameplay, fixed bugs and iterated.
Original post ↗Ryan Carson uses Devin's Test App feature to record and narrate feature clicks, fix bugs, and merge PRs with a !land command.
Original post ↗A developer used Claude Code as a pair programmer for provisioning, database seeding, SSL, DNS cutover and production debugging, finishing in about four hours with zero downtime by his account.
Original post ↗Google's Big Sleep agent reviewed recent SQLite commits starting from a patched bug and found an exploitable flaw, which was fixed before release.
Original post ↗Cognition staff assign Devin work in Slack or Linear; it posts PRs, fixes bugs flagged by review bots and works CI and lint failures until checks pass.
Original post ↗A reviewer used the Claude Chrome extension for Amazon shopping, building an interactive model timeline from one prompt, and translating and summarizing pages; one screenshot bug appeared.
Original post ↗Cursor's support team uses Cursor with MCP to pull Datadog logs, search past cases, and draft a Linear bug ticket and customer reply.
Original post ↗Adam uses Claude Code via an MCP server to search entities, read automation traces and build and debug automations like a shed motion light.
Original post ↗OpenAI's internal OpenClaw agent Androidclaw uses Git, GitHub and logs to investigate broken builds, trace problems to incidents, and sometimes prepare and merge fixes.
Original post ↗Hermes Agent handled payables, bank import, bill posting and Stripe reconciliation alongside a human accountant, but needed 7 corrections including a tax-amount bug.
Original post ↗Coding agents can reproduce many bugs from a report or stack trace, write a fix and open a pull request with a test. Keep them on a branch and require review, since a plausible-looking fix can still miss the real cause.
Across the 41 published workflows on this page, the most used tools are Claude Code (10), Devin (5) and Cursor (4). Each use case lists every tool it needs and the role it plays.
Open any use case to see the tools, the step-by-step setup and example prompts you can copy. Start with a beginner workflow, run it on a small batch and review the output before you automate it fully.