delta-v. Dev notes on AI agents, coding tools and Python. Experiments, frameworks and tools I build while learning.
33 posts · 2 categories · 112 tags · latest Jul 25, 2026
45 repos tracked · 4262.1k ★ combined · top obra/superpowers (296.5k ★)
Coding agents default to the same purple gradient, the same Inter, the same three centred cards. Three open-source design skills fight that — with very different bets on how. What each actually does, how they compare, and why the one benchmark that exists is circular.
Humanizer is a portable, open-source agent skill (Claude Code and any skill-aware harness) that rewrites text to remove the 33 tells of AI writing — em-dash overuse, rule-of-three, 'let's dive in', significance inflation — based on Wikipedia's Signs of AI writing. What it is, how it works, the reviews, and the honest pros, cons, and ethics.
Understand Anything is an open-source Claude Code plugin that runs a multi-agent, tree-sitter + LLM pipeline over your repo and builds an interactive knowledge graph — files, functions, dependencies, domains — you can search, tour, and ask questions about. What it is, how it works, real reviews, and the honest pros and cons.
DevSwarm is a proprietary desktop IDE built on VS Code that runs multiple AI coding agents in parallel, each in its own isolated Git worktree, with cross-agent orchestration and Jira/GitHub review. What it is, how it works, the honest caveats, and how it stacks up against Herdr and tmux.
Coding agents default to the same purple gradient, the same Inter, the same three centred cards. Three open-source design skills fight that — with very different bets on how. What each actually does, how they compare, and why the one benchmark that exists is circular.
Humanizer is a portable, open-source agent skill (Claude Code and any skill-aware harness) that rewrites text to remove the 33 tells of AI writing — em-dash overuse, rule-of-three, 'let's dive in', significance inflation — based on Wikipedia's Signs of AI writing. What it is, how it works, the reviews, and the honest pros, cons, and ethics.
Understand Anything is an open-source Claude Code plugin that runs a multi-agent, tree-sitter + LLM pipeline over your repo and builds an interactive knowledge graph — files, functions, dependencies, domains — you can search, tour, and ask questions about. What it is, how it works, real reviews, and the honest pros and cons.
DevSwarm is a proprietary desktop IDE built on VS Code that runs multiple AI coding agents in parallel, each in its own isolated Git worktree, with cross-agent orchestration and Jira/GitHub review. What it is, how it works, the honest caveats, and how it stacks up against Herdr and tmux.
Herdr is an agent multiplexer — tmux for coding agents. Real terminal panes, blocked/working/done state at a glance, persistence on detach, SSH from your phone, and an API agents can drive. What it is, how it works, the pros and cons versus tmux, and what early users say.
A practical 0-to-hero guide to Obsidian, the local-first markdown knowledge base. What it is and why it matters, how the vault actually works, the essential power-user plugins (Dataview, Templater, Bases, Tasks), the tips real power users swear by, honest pricing, and where it falls short.
Coding agents can edit your repo but choke the moment you ask them to read a tweet, a YouTube transcript, or a Reddit thread. Agent Reach is a capability layer that picks, installs, health-checks and routes the best free backend for each platform — what it is, how the backend-routing works, and the real risks (account bans, vibe-coded code, China-centric coverage).
Headroom sits between your AI agent and the model, compressing tool outputs, logs, RAG chunks, files and history before they cost you tokens — reversibly, and locally. What it is, how the router-plus-compressors design works, what its self-reported benchmarks actually show, and the honest costs.
agency-agents is an open-source collection of 218 specialized subagent 'personalities' you can install into Claude Code and a dozen other tools. What it is, how to install and use it, how it compares to writing your own, and the honest caveats.
Career-Ops is an open-source, agentic job-search pipeline that scores offers, tailors your CV, scans portals, and tracks everything — locally, with you in the loop. What it is, how to run it, how it compares to the alternatives, and the honest caveats.
find-skills turns 'I wish my agent could do X' into 'find and install a skill for X' — automatically. How it works with the skills CLI and the skills.sh registry, how to use it, and the trust caveats of letting an agent install its own capabilities.
Ponytail is a one-file 'lazy senior dev' skill that makes coding agents stop and pick the simplest solution that works — cutting code by ~54% (and tokens ~22%) on a real agentic benchmark. What it is, how the ladder works, what trusted reviewers found, and the honest caveats.
Ralph is a brutally simple technique — loop the same prompt into a coding agent until the task is done. What it is, how the two main implementations (Anthropic's ralph-wiggum plugin and snarktank/ralph) actually differ, the anecdotal numbers behind the hype, and an honest critical review of where it breaks.
What agent skills are, how Claude Code loads them (and what they cost in context), the real risks, and a lean, honest stack for code review, security, quality and multi-agent work — with install/remove/update for each.
CodeBurn is a local TUI dashboard that tracks token usage, cost, and performance across Claude Code, Codex, Cursor and ~18 other AI coding tools. What it is, how to read the dashboard, how it compares to ccusage, and its pros and cons.
Graphify is a Claude Code skill that compiles code, docs, papers and images into a queryable knowledge graph — so an agent answers from a tiny subgraph instead of re-reading raw files. How it works, how it saves tokens, the real benchmarks, and how it compares to vector RAG and the LLM Wiki.
LiteLLM is an open-source AI gateway that gives you a single OpenAI-format interface to 100+ LLM providers — as a Python SDK or a self-hosted proxy. What it is, how it works, two real use cases, how it compares, its performance, and its honest pros and cons.
OpenSpec is an open-source, lightweight spec-driven-development tool that adds a spec layer so humans and AI agree on what to build before code. What it is, how the propose → apply → archive workflow and delta specs work, how it compares to Spec Kit and Kiro, and its honest tradeoffs.
pi is a tiny, aggressively-extensible terminal coding harness. What it is, how to install and use it, how it compares to opencode and Claude Code, and what the benchmarks actually say.
Spec Kitty is an open-source Python CLI that turns product intent into a repo-native, spec-driven workflow — spec, plan, tasks, review, accept, merge — with isolated git worktrees for parallel agents, a local kanban dashboard, and governance gates. What it is, how it works, use cases, and where it fits versus lighter tools.
Two open-source takes on spec-driven development for AI coding agents — Spec Kitty's governed software factory versus OpenSpec's lightweight spec layer. The real differences, pros and cons, examples, and when to reach for each.
A practical look at LangChain — what it's for, how to install and use it, what you can build, whether you need to deploy it, how it stacks up against Google ADK and other agent frameworks, and its real downsides.
A community-built, Apache-2.0 library of 754 structured cybersecurity skills that give an AI agent senior-analyst playbooks — mapped to MITRE ATT&CK, NIST CSF 2.0, ATLAS, D3FEND and NIST AI RMF. What it is, how the skills are built, defensive use cases, how it compares, and the honest tradeoffs (including its dual-use responsibility).
Addy Osmani's Agent Skills packs the workflows, quality gates, and Google engineering practices that senior engineers use into markdown skills an AI coding agent follows across the whole development lifecycle. What it is, how it works, and the honest tradeoffs.
mattpocock/skills is Matt Pocock's everyday .claude skills folder, published — small, composable agent skills that fix the common failure modes of coding agents (misalignment, verbosity, broken code, ball-of-mud architecture) without taking over your process. What it is, how it works, the standout grilling and TDD skills, and the honest tradeoffs.
A single CLAUDE.md file, derived from Andrej Karpathy's observations on LLM coding pitfalls, that distills four principles — think before coding, simplicity first, surgical changes, goal-driven execution — to make Claude Code (and Cursor) behave. What it is, how it works, and an honest read on its viral metrics.
UI UX Pro Max is an open-source AI skill that gives a coding agent a design brain — a reasoning engine that turns a one-line prompt into a complete design system with styles, palettes, typography, and anti-patterns. What it is, how it works, what it produces, and the honest tradeoffs.
A plain-English explainer of the AI harness — the code around a model that lets it use tools, take steps, and get work done. What it is, why it matters, how it works, and the two senses of the word (agent harness vs evaluation harness).
awesome-claude-code is a selectively curated 'awesome list' of skills, agents, hooks, slash commands, status lines, CLAUDE.md files and tooling for Claude Code. What it is, how it's organized and curated, the standout categories, how it compares to bigger aggregators, and the honest tradeoffs.
Claude-Mem is an open-source plugin that gives Claude Code memory across sessions by capturing what the agent does, compressing it into structured observations, and re-injecting the relevant slice next time. What it is, how it works, real reviews, and its honest tradeoffs.
Andrej Karpathy's LLM Wiki pattern — let an LLM build and maintain a persistent, interlinked wiki from your sources, so knowledge compounds instead of being re-derived on every query. What it is, how it works, how it compares to RAG, and the token economics.
A Claude Code skill that makes the agent talk like a caveman — cutting output tokens by up to ~75% while keeping full technical accuracy. Why, how it works, how to install it, and real before/after numbers.
Superpowers is an open-source plugin that gives a coding agent a full development methodology — brainstorm, plan, then build with TDD and subagent review. What it is, how it works (and why it stays token-light), real opinions, use cases, and its pros and cons.
TradingAgents: Multi-Agents LLM Financial Trading Framework
基于多智能体LLM的中文金融交易框架 - TradingAgents中文增强版
Download market data from Yahoo! Finance's API
An agentic skills framework & software development methodology that works.
→ related post: Superpowers: A Software Development Methodology for Your Coding Agent
Skills for Real Engineers. Straight from my .agents directory.
→ related post: Matt Pocock's Skills for Real Engineers: Small, Composable Agent Skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
→ related post: andrej-karpathy-skills: A 70-Line CLAUDE.md That Stops Agents Overengineering
Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
→ related post: The Agency: A 218-Strong Library of Specialized AI Subagents
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
→ related post: Ponytail: The Best Code Is the Code Your Agent Never Writes
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
The agent engineering platform.
→ related post: LangChain: What It Is, How It Compares, and When to Reach for It
An AI skill that provides design intelligence for building professional UI/UX across multiple platforms.
→ related post: UI UX Pro Max: A Design-Intelligence Skill for Your Coding Agent
AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI
→ related post: What Is a Harness? The Scaffolding That Turns a Model Into an Agent
🪨 why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
→ related post: Caveman: Why Use Many Token When Few Do Trick
Production-grade engineering skills for AI coding agents.
→ related post: Agent Skills: Bolting Senior-Engineer Discipline Onto Your Coding Agent
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
→ related post: Claude-Mem: Persistent Memory for Your Coding Agent
Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
→ related post: Agent Reach: One Command to Give Your Coding Agent the Open Internet
Taste-Skill - gives your AI good taste. stops the AI from generating boring, generic slop
→ related post: Three Skills That Teach Your Agent Taste: Taste Skill, Impeccable, Hallmark
A high-throughput and memory-efficient inference and serving engine for LLMs
Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
→ related post: Understand Anything: Turn a Codebase Into a Knowledge Graph You Can Explore
The design language that makes your AI harness better at design.
→ related post: Three Skills That Teach Your Agent Taste: Taste Skill, Impeccable, Hallmark
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, FLUX and more.
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
→ related post: Headroom: A Compression Layer That Shrinks Everything Your Agent Reads
Spec-driven development (SDD) for AI coding assistants.
→ related post: Spec Kitty vs OpenSpec: Which Spec-Driven Tool Should You Use?
The Memory Layer for AI Agents - Drop-in memory infrastructure for AI agents and apps. Context that persists. Built for production.
The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]
A hand-picked collection of the finest of resources for the most awesome of agents, Claude Code, the undisputed champion of coding companions, from the unstoppable team at Anthropic PBC. A delectable showcase of top tier skills, ambidextrous agents, scintillating status lines, top notch developer tooling, and also we have plugins
→ related post: Awesome Claude Code: A Curated Directory for Your Claude Code Workflow
Agent skill that removes signs of AI-generated writing from text
→ related post: Humanizer: A Skill That Strips the AI Tells Out of Your Writing
LlamaIndex is the document processing platform for AI
817 structured cybersecurity skills for AI agents · Mapped to 6 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND, NIST AI RMF & MITRE F3 (Fight Fraud) · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platforms · 29 security domains · Apache 2.0
→ related post: Anthropic Cybersecurity Skills: A Structured Security Playbook for AI Agents
The open agent skills tool - npx skills
→ related post: Powering Up Claude Code: A Practical Guide to Dev Skills
Anti-AI-slop design skill for Claude Code, Cursor, and Codex.
→ related post: Three Skills That Teach Your Agent Taste: Taste Skill, Impeccable, Hallmark
Ralph is an autonomous AI agent loop that runs repeatedly until all PRD items are complete.
→ related post: Ralph: Put a Coding Agent in a While Loop and Walk Away
Free, local tool to track AI coding token usage and cost across 37 tools and agents (Claude Code, Cursor, Codex, Gemini and more), by model, project, and task. npx codeburn
→ related post: CodeBurn: See Where Your AI Coding Tokens Actually Go
Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
Flexible and powerful data analysis / manipulation library for Python, providing labeled data structures similar to R data.frame objects, statistical functions, and much more
LLM training in simple, raw C/CUDA
An agentic skills framework & software development methodology that works.
→ related post: Superpowers: A Software Development Methodology for Your Coding Agent
Skills for Real Engineers. Straight from my .agents directory.
→ related post: Matt Pocock's Skills for Real Engineers: Small, Composable Agent Skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
→ related post: andrej-karpathy-skills: A 70-Line CLAUDE.md That Stops Agents Overengineering
Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
→ related post: The Agency: A 218-Strong Library of Specialized AI Subagents
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
→ related post: Ponytail: The Best Code Is the Code Your Agent Never Writes
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
The agent engineering platform.
→ related post: LangChain: What It Is, How It Compares, and When to Reach for It
An AI skill that provides design intelligence for building professional UI/UX across multiple platforms.
→ related post: UI UX Pro Max: A Design-Intelligence Skill for Your Coding Agent
AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI
→ related post: What Is a Harness? The Scaffolding That Turns a Model Into an Agent
🪨 why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
→ related post: Caveman: Why Use Many Token When Few Do Trick
TradingAgents: Multi-Agents LLM Financial Trading Framework
Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
Production-grade engineering skills for AI coding agents.
→ related post: Agent Skills: Bolting Senior-Engineer Discipline Onto Your Coding Agent
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
→ related post: Claude-Mem: Persistent Memory for Your Coding Agent
Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
→ related post: Agent Reach: One Command to Give Your Coding Agent the Open Internet
Taste-Skill - gives your AI good taste. stops the AI from generating boring, generic slop
→ related post: Three Skills That Teach Your Agent Taste: Taste Skill, Impeccable, Hallmark
A high-throughput and memory-efficient inference and serving engine for LLMs
Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
→ related post: Understand Anything: Turn a Codebase Into a Knowledge Graph You Can Explore
The design language that makes your AI harness better at design.
→ related post: Three Skills That Teach Your Agent Taste: Taste Skill, Impeccable, Hallmark
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, FLUX and more.
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
→ related post: Headroom: A Compression Layer That Shrinks Everything Your Agent Reads
Spec-driven development (SDD) for AI coding assistants.
→ related post: Spec Kitty vs OpenSpec: Which Spec-Driven Tool Should You Use?
The Memory Layer for AI Agents - Drop-in memory infrastructure for AI agents and apps. Context that persists. Built for production.
The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]
A hand-picked collection of the finest of resources for the most awesome of agents, Claude Code, the undisputed champion of coding companions, from the unstoppable team at Anthropic PBC. A delectable showcase of top tier skills, ambidextrous agents, scintillating status lines, top notch developer tooling, and also we have plugins
→ related post: Awesome Claude Code: A Curated Directory for Your Claude Code Workflow
Agent skill that removes signs of AI-generated writing from text
→ related post: Humanizer: A Skill That Strips the AI Tells Out of Your Writing
LlamaIndex is the document processing platform for AI
Flexible and powerful data analysis / manipulation library for Python, providing labeled data structures similar to R data.frame objects, statistical functions, and much more
817 structured cybersecurity skills for AI agents · Mapped to 6 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND, NIST AI RMF & MITRE F3 (Fight Fraud) · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platforms · 29 security domains · Apache 2.0
→ related post: Anthropic Cybersecurity Skills: A Structured Security Playbook for AI Agents
The open agent skills tool - npx skills
→ related post: Powering Up Claude Code: A Practical Guide to Dev Skills
基于多智能体LLM的中文金融交易框架 - TradingAgents中文增强版
LLM training in simple, raw C/CUDA
Anti-AI-slop design skill for Claude Code, Cursor, and Codex.
→ related post: Three Skills That Teach Your Agent Taste: Taste Skill, Impeccable, Hallmark
Download market data from Yahoo! Finance's API
Ralph is an autonomous AI agent loop that runs repeatedly until all PRD items are complete.
→ related post: Ralph: Put a Coding Agent in a While Loop and Walk Away
Free, local tool to track AI coding token usage and cost across 37 tools and agents (Claude Code, Cursor, Codex, Gemini and more), by model, project, and task. npx codeburn
→ related post: CodeBurn: See Where Your AI Coding Tokens Actually Go