AGENTS.md

Headlines · 4

1. 4d
   Geoffrey Huntley on using pre-commit hooks to keep coding agents in line

   In a quote highlighted on Simon Willison's blog, Geoffrey Huntley describes how he keeps a coding agent using his diagnostics library: if the agent bypasses it, he writes a pre-commit hook that checks the git diff for direct calls to observability modules that skip the diagnostics module. The hook fails the commit and injects a prompt telling the agent to use the diagnostics library, which he says turns the outcome into "ten out of ten." He notes that introspecting and updating AGENTS.md instead still leaves a delta that will drift, and the quote sits alongside his view that nobody should type code by hand anymore.

   Simon Willison's Weblog · 🔥 0
2. 7d
   GOAI 2026: AI open source shifts from code availability to reusable capability

   This InfoQ analysis argues that the value of AI open source is shifting from "what is open" to whether the released capability can be verified, reused and continuously iterated: downloadable weights do not mean a system can be reproduced, since AI projects also depend on data pipelines, prompts, tool interfaces, runtime environments and hardware, while agent systems add task planning, state management and permission control. It uses the 2026 Global Open-source AI competition (GOAI) as a sample, reporting 91 countries and regions, over 14,000 developers, 2,999 valid preliminary entries and 70 finalists, and cites the Linux Foundation's December 2025 launch of the Agentic AI Foundation (covering MCP, goose and AGENTS.md), OSI's Open Source AI Definition 1.0, Anthropic's technical article on evaluating agents through full execution traces, and workflow tracing in the OpenAI Agents SDK. It also argues that reusable open assets include data versions, evaluation scripts, run traces, experiment records and simulation environments, and describes Hangzhou's use of GOAI to link global developers with industry resources.

   InfoQ 中文 AI&大模型 · 🔥 0
3. 7d
   Claude Code will support AGENTS.md, as Anthropic weighs a future without CLAUDE.md

   Claude Code team member Thariq Shihipar said on the Latent Space podcast that Claude Code will support AGENTS.md, the cross-tool instruction file it had previously declined to adopt, a stance that drew public criticism. He argued that as models improve, static instruction files like CLAUDE.md may eventually disappear, and that new projects can skip one entirely, adding rules only after seeing the same failure pattern repeatedly. He also described Claude Mods for customizing the whole harness (execution logic and UI), artifacts as a persistent generative interface, and Claude Tag and Projects for multi-person work.

   InfoQ 中文 AI&大模型 · 🔥 0
4. 7d
   SwiftUI Pro agent skill launches for Claude Code, Codex and more

   Developer Paul Hudson (Hacking with Swift) published SwiftUI Pro on GitHub, an agent skill that helps AI coding assistants such as Claude Code, Codex, Gemini and Cursor write more modern, simpler SwiftUI. The skill builds on his existing AGENTS.md work and targets mistakes LLMs actually make — deprecated API usage, VoiceOver/accessibility problems and surprising performance issues — covering navigation, layout, animations and state management. It installs via npx skills add or the Claude Code plugin marketplace and is triggered with /swiftui-pro in Claude Code or $swiftui-pro in Codex; it is MIT-licensed, with sibling skills including SwiftData Pro, Swift Concurrency Pro and Swift Testing Pro.

   GitHub Trending（每日） · 🔥 0

Experience and discussion from the community

Share my AGENTS.md experienceAsk about AGENTS.md