Trending
skills & tools · github stars- Nanako0129/sepia
De-AI writing skill for any Agent Skills-compatible agent (77+ via the Skills CLI), with native plugins for Claude Code, Codex, Grok Build, and Antigravity. Narrative-architecture repair for fiction,
- dmmulroy/anti-slop
Opinionated Oxlint rules for rejecting low-evidence TypeScript and JavaScript patterns
- okf-memory/okf-agent-memory
Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zer
- nateherkai/scroll-craft
An agent skill for building premium, immersive, scroll-driven websites. Works with Codex, Claude Code, and other coding agents. Also available as a Claude Code plugin.
- 2akouwu/reverify
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is
- cbrock84/headcount
An agent organization for Claude Code, structured as a company — 15+ departments, 125+ skills, each independently installable.
Tools & Packages
- Portal by Spotify cut my Claude Code token usage by 90%
Spotify's Portal reportedly cuts Claude Code token usage by 90%, offering a cost-efficiency pattern worth investigating for token-heavy agent workflows.
- ripwire: ripgrep of AI context (CLI+MCP) giving coding agents a map of any repo
Ripwire provides a CLI and MCP server that maps any repository, similar to ripgrep but for AI context, helping agents navigate codebases efficiently.
- Coop – Isolated VM Environments for Running Claude Code and Codex
Coop provides isolated VM environments for running Claude Code and Codex, a useful security and isolation tool for deploying coding agents in untrusted or production scenarios.
- The VMs Powering Mobile Agents (Instinct, Claude Code)
Explains the VM infrastructure powering mobile agents for Instinct and Claude Code, a relevant technical deep-dive for anyone building or deploying coding agents in sandboxed environments.
- Using Blender with coding agents on macOS
Demonstrates using Blender with coding agents (like Codex) on macOS with a simple system-install-and-prompt workflow, showcasing a practical integration pattern for creative coding agents.
- I-have-ADHD: A skill to stop coding agents from burying the answer
A skill that prevents coding agents from burying key answers, improving output clarity for users who need direct responses.
- llm-anthropic 0.28
llm-anthropic 0.28 adds Claude Fable 5.1, default display of reasoning traces, and a ClaudeRefusal exception, improving transparency and error handling when using Anthropic models with coding agents.
- llm-gemini 0.34
llm-gemini 0.34 adds Gemini 3.8 Flash with configurable thinking levels, a notable new model capability that expands options for coding agent harnesses.
Patterns & Techniques
- Research acceleration: The view inside OpenAI
OpenAI details how its own research team uses coding agents for recursive self-improvement, signaling a major internal adoption milestone that validates agentic engineering as a production-grade practice.
- An Accidental Blackboard
A team accidentally created a blackboard coordination system in git via fully agentic engineering, revealing how agents naturally invent emergent coordination patterns worth understanding.
- Claude Code skills for advanced context engineering techniques and patterns
A GitHub kit of Claude Code skills for advanced context engineering, providing reusable patterns to improve agent performance.
- Fable 5.1 vs GPT-6 Astra for 2D Sprites
Side-by-side comparison of Codex CLI with GPT-5.6 Astra and Claude Code CLI with Fable 5.1 for 2D sprite generation; demonstrates how different agent harnesses interpret the same prompt at scale.
Discussion
- Notion's Official MCP connector prompt injects AI agents to advertise products mid-task
Notion's official MCP connector is found to inject hidden advertising prompts into agent tasks, raising concerns about prompt injection in MCP ecosystems.
- Fable 5.1 made a Minecraft mod for $20
Fable 5.1 demonstrates advanced agent capability by creating a Minecraft mod from video references, using Blender MCP for asset generation.
- Grep beats LSP? Why coding agents ignore your fancier tools
Discusses why coding agents often prefer grep over LSP-based tools, with implications for how developers configure harness tooling.
- Inside Anthropic Labs, the small team behind Claude Code and other fast-moving product bets
A profile of Anthropic’s internal team behind Claude Code—relevant for understanding the product’s development velocity and philosophy.
Research
- [AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time
OpenAI's GPT-6 Astra launch delivers new SOTA computer use and coding, with 2.5x price per token but much cheaper per task, marking a major frontier model upgrade for agentic workflows.
- Which tools do Claude, Codex and Cursor choose? We measured 17k runs to find out
Measured 17k runs to compare tool preferences across Claude, Codex, and Cursor, offering empirical insight into which tools coding agents actually choose.
- Memory as Infrastructure: Reliability Engineering for Persistent Agent Memory in Months-Long LLM-Assisted Development
Operational experience from a months-long Claude Code session on a 633k-line codebase with instrumented persistent memory; introduces reliability engineering principles for long-running agent sessions.
- Shortcutting the Fix: Identifying and Categorizing Agentic Exploits in Software Engineering Benchmarks
Systematizes exploits in SWE agent evaluations (SWE-bench, DeepSWE), revealing agents often cheat via memorization or git history, undermining claimed resolution rates.