Topic
4 entries
AI agents are writing code faster than humans can review, creating a bottleneck in software development workflows. This article proposes four strategies used by PostHog to leverage AI for offloading code review and adjacent tasks, allowing human developers to review less code and focus on complex issues. These methods include making agents review each other's work, delegating PR babysitting, implementing an auto-stamper for low-risk PRs, and verifying agent-generated code through observation rather than reasoning. By adopting these approaches, teams can maintain quality, accelerate shipping, and reduce developer fatigue.
The `qa-swarm` is an AI-powered GitHub bot skill designed to provide comprehensive, multi-perspective Pull Request (PR) reviews. It orchestrates four independent AI agents—qa-team, paul-reviewer, xp-reviewer, and security-audit—to run in parallel, each bringing a unique review perspective. The skill synthesizes their findings, deduplicates them, and posts them as inline GitHub PR comments and an upserted summary comment with a calculated verdict. This ensures thorough code review while clearly marking all automated feedback.
Ponytail is an AI agent skill designed to make large language models (LLMs) think like a "lazy senior dev," prioritizing existing solutions and minimal code to prevent over-engineering. It guides agents through a decision-making ladder, focusing on reuse of standard libraries, native features, and installed dependencies before writing new code. This approach leads to significant reductions in code volume (up to 94%), cost (20%), and execution time (27%) in real-world benchmarks, while maintaining code safety. The tool offers plugins and integration for various AI agents and developer environments like Claude Code, GitHub Copilot, and Devin CLI.
This GitHub repository provides a comprehensive curated list of "Claude Skills," which are specialized, reusable folders containing instructions, scripts, and resources for customizing Claude AI workflows, particularly Claude Code. Skills function via a progressive disclosure architecture, loading only metadata initially for efficiency, and are dynamically discovered by Claude for relevant tasks. The resource details skill creation, installation via web, CLI, or API, showcases numerous official examples.
