Trending
skills & tools · github stars- img2threejs/img2threejs
Rebuild the object in a reference image as a code-only, procedural, quality-gated, animation-ready Three.js model. Token-efficient image-to-3D.
- Vincentwei1021/video-shotcraft
AI video skill for Claude Code & Codex — cinematic product videos with Remotion: 106 shot recipe cards, 161 motion previews, a production-ready template
- DietrichGebert/ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
- 0xwilliamortiz/openclaude-improved
runs anywhere. uses anything
- makecindy/cindy
Consider it done. The open-source AI agent that works out of the box · 想到，就能做到。开源、开箱即用的 AI Agent。
- lidge-jun/opencodex
Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code
Tools & Packages
- "We removed over 80% of Claude Code's system prompt for Opus 5 and Fable 5"
Anthropic removed over 80% of Claude Code's system prompt for Opus 5 and Fable 5, a significant optimization for coding agent efficiency and maintainability.
- Claude Code has a hardcoded instruction telling Opus 5 not to use subagents
Claude Code contains a hardcoded instruction preventing Opus 5 from using subagents, revealing deliberate constraints in agent orchestration design.
- Oh-my-pi: A coding agent with the IDE wired in
Oh-my-pi integrates an IDE into a coding agent flow, potentially enabling tighter code-editing loops.
- How well do you use Claude Code?
Promptster.ai offers analytics and coaching for coding agent (Claude Code/Codex) usage, helping teams optimize spend and workflow effectiveness.
- CORVUS: Context Optimization and Reduction Via Underlying Synchronization for LLM Coding Agents
提出CORVUS方法，优化编码代理轨迹中的上下文，避免文件快照过时，提高决策准确性，是改进代理上下文管理的关键技术。
Patterns & Techniques
- feat(managed_agents): add subagent live-streaming cookbook
New cookbook demonstrates live-streaming coordinator and subagents, covering per-thread event deltas, per-agent model effort, and versionless agent updates.
- Execution-Grounded Security Testing for Coding Agents in Software Engineering Pipelines
提出针对编码代理在系统操作中的安全测试框架，识别持久化修改和权限滥用风险，是编码代理安全的重要方法。
- The harness is all you need (mostly)
A practical GitHub Copilot workflow for prototyping, planning, implementing, and reviewing software without chasing every new AI tool.
Discussion
- A Fireside Chat with Cat and Thariq from the Claude Code team
Fireside chat with the Claude Code team covering agent security, evals, tool design, and internal usage patterns—essential viewing for anyone building with coding agents.
- Fragments: July 21
Thoughtworks retreat finds that verification, not code generation, is now the bottleneck, and 'harness engineering' is emerging as a distinct discipline—critical insight for coding agent development.
- Quoting Thomas Ptacek
Thomas Ptacek argues that an open-weights model with a pentest harness could compromise most networks, emphasizing the need for robust sandboxing when building coding agents.
- Copilot vs. raw API access: What are you actually paying for?
Compares Copilot's pricing with raw API access, highlighting the value of coding workflow and policy around the model.
Research
- The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages
Research showing significant token cost differences across programming languages for coding agents, with analysis of why some languages are more token-efficient.
- StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents
StateAct is a code-first multi-agent harness that operates on program state directly rather than screenshots, improving long-horizon computer-use tasks.
- Quoting Boris Cherny
Anthropic's Opus 5 shows significant improvement in prompt injection resistance, a crucial security property for coding agents that execute code based on LLM output.