RTK: Open-Source CLI Proxy for AI Coding Agents

Claude Code's source was accidentally published to npm. So I studied every prompt in the codebase using claude. Here's what I found and I'm open-sourcing all of it. Claude Code uses 26 distinct prompts to function: > 1 system prompt (identity, safety, code style, tool routing) > 11 tool prompts (shell, file ops, search, web, planning) > 5 agent prompts (explorer, architect, verifier, docs, general) > 4 memory prompts (summarization, session notes, extraction) > 1 coordinator prompt (multi-agent orchestration) > 4 utility prompts (titles, recaps, suggestions) The patterns that stood out: 1. Anti-over-engineering rules: "don't add features beyond what was asked" 2. Tiered risk assessment : freely edit files, but confirm before force-pushing 3. Adversarial verification : a dedicated agent whose job is to TRY TO BREAK the implementation 4. Memory compression : 9-section summarization that preserves every user message 5. Never delegate understanding : "write prompts that prove you understood" I have rewritten every prompt from scratch for legal compliance. Same behavioral intent wihout verbatim copying text. The repo includes: > Every prompt, ready to copy into your own agent > 9 pattern analyses with commentary > 3 claude skills you can drop in today > MIT licensed you can fork and reuse as it is. If you're building AI coding agents, this will save you months of prompt engineering. Link: https://lnkd.in/gNizmf6T #PromptEngineering #ClaudeCode #AI #AIAgents #LLM #OpenSource

If you're building AI agents and haven't seen this yet : it's every prompt Claude Code uses, rewritten and open-sourced. The verification agent pattern alone is worth the click.

A lot of token spend with large language models (LLMs) is not your prompt. If you use an LLM while debugging from the terminal, this adds up fast. It is the terminal noise you paste into the context window: `git diff`, test logs, `grep` output. One pragmatic idea: treat command output (stdout) like an API surface. RTK (https://lnkd.in/dwt3Sycp) positions itself as a local CLI proxy. It runs your command and filters or compresses stdout before it reaches the model context. The RTK README claims 60–90% token reduction. How it gets there (per command type): - Smart filtering (remove noise) - Grouping (aggregate similar items) - Truncation (keep relevant context, cut redundancy) - Deduplication (collapse repeats with counts) Adoption detail I care about: It can auto-rewrite commands so `git status` becomes `rtk git status` via a PreToolUse hook (a command rewriter). The model never sees the rewrite. Safety net: on failure, it can `tee` the full unfiltered output to a saved log and print the path. Privacy note: telemetry is enabled by default (anonymous aggregate metrics, once per day). Opt out with `RTK_TELEMETRY_DISABLED=1` or setting it at `~/.config/rtk/config.toml`. What would stop you from using stdout filtering: correctness, debugging, or privacy? #llm #ai #developer #token #rtk #proxy #opensource

AI coding agents are incredible, but they all suffer from the same fatal flaw: Attention Dilution. If you’ve ever unleashed an agent on a massive, N years old legacy codebase, you know what happens. They use standard cat or grep, swallow thousands of lines of boilerplate just to change one function, burn through tokens, and eventually hallucinate. Vector RAG isn't the answer for code either,it loses it and goes bananas. To solve this, I’ve been building OptiVault: an open-source Static Context Compiler and Model Context Protocol (MCP) server. Instead of dumping raw code into the context window, OptiVault intercepts it. A Zero-Fat (Keto ?) AI Developer workflow: AST-Driven Semantic Routing: Powered by Tree-sitter, OptiVault extracts deterministic function signatures and dependency skeletons. LLMs drill down hierarchically without the bloat. Obsidian Dual-Compatibility: The shadow context acts as both an AI index AND a human-readable Obsidian Knowledge Graph. The Autopilot Loop: OptiVault generates a CLAUDE.md that teaches agents to use MCP tools, sync their own context after edits, and strictly following best agent coding principles. GitHub: https://lnkd.in/eiZpsG6g #AI #SoftwareEngineering #Claude #MCP #OpenSource #Token #Obsidian

trying to maximize #OptiVault, I think claude is making tokens go on a stress management for longer tasks, its still a cat-and-mouse fight. But I couldn't sleep much, couldn't let my brain go with everything going on, so I made OptiVault go on a diet with Zero-Fat Navigation (query_graph, concepts merge, strict protocol, real benchmark). after you install it using npm and link it to your claude, you can test on your project: cd ~/some-other-repo optivault init . # indexes the repo, creates _optivault/ + CLAUDE.md claude mcp add optivault optivault -- mcp \ --vault "$(pwd)/_optivault" \ --source "$(pwd)" Then open #ClaudeCode in that repo — it'll pick up the new CLAUDE.md protocol and all 6 MCP tools including the new query_graph.

If you’re using Claude Code, this is the one repository you actually need to bookmark. The everything-claude-code project (by affaan-m) has officially crossed 100k stars this month, evolving from a hackathon winner into the definitive "performance engine" for AI agents. 🚀 Why it’s the gold standard: 28+ Specialized Agents: Pre-configured roles for TDD, Security Audits, and Research. 120+ Battle-Tested Skills: Modular powers like automated refactoring and complex dependency mapping. AgentShield Security: Built-in hooks to prevent secret leaks and "hallucinated" git commands. Memory Persistence: Finally, a way to keep context across sessions without hitting the token ceiling. Whether you're on Claude Code, Cursor, or Codex, this repo provides the structural "brains" to make them behave like senior engineers rather than just chatbots. 🔗 Check it out: https://lnkd.in/gbRVG2sm #ClaudeCode #AI #OpenSource #SoftwareEngineering #AgenticWorkflows

I built something I haven't seen anywhere else: a skill that makes AI coding agents look 👁️ forward 🔮 after they write code. Every agent skill today is about doing the task. Write the feature. Fix the bug. None of them ask: what did that just break? Foresight is the first post-implementation lookahead skill. It runs automatically at the end of any implementation turn and produces a compact block — under 120 words — covering: → What will crash immediately in the existing codebase   → Silent edge cases that pass tests but fail in production   → Security surfaces opened by the change   → Pattern inconsistencies that become tech debt   → The exact question the user is about to ask next  That last one is the most underrated. Not "you should add tests" — but "why is canVote always returning false for users over 18?" Specific. Grounded. Saves a full turn. I ran a 60-case benchmark across bug fixes, refactors, schema changes, and architecture shifts: - Without the skill: 32% of risks surfaced - With foresight: 100% across all 60 cases 🔥 - Net token economics: the skill adds ~35 words per turn and avoids ~77 words 👏 of follow-up prompting. It pays for itself on the first miss it prevents. The gap that surprised me: baseline catches the obvious crash almost every time. It almost never catches the silent edge case, the pattern break, or the security implication. And it never predicts what you're about to ask. The skill is live. The benchmark is open. The methodology doc covers what we tested, how we tested it without an API key, and the honest limitations of the approach. https://lnkd.in/gwZnyDi9 If you're building with Claude Code, this is worth trying. An agent that looks forward after writing code is a fundamentally different tool. Available in the terminal close to you: npx skills add tombrewsviews/foresight-skill

BREAKING: The largest collection of AI coding skills just hit 31,000+ stars. 1381+ skills. One repo. Works everywhere. → Claude Code → Gemini CLI → Codex CLI → Cursor → GitHub Copilot → OpenCode → Antigravity IDE → AdaL CLI What are skills? AI agents are smart but generic. They don't know your deployment protocol. They don't know your company's architecture patterns. They don't know AWS CloudFormation syntax. Skills are small markdown files that teach them. One skill = one capability. Perfectly executed. Every time. This repo has : → Architecture (system design, ADRs, C4) → Security (AppSec, pentesting, compliance) → DevOps (Docker, AWS, Vercel, CI/CD) → Data & AI (RAG, agents, LangGraph) → Testing (TDD, QA workflows) → Business (SEO, pricing, copywriting) Install once: npx antigravity-awesome-skills Then: "Use @brainstorming to plan a SaaS MVP." "Run @lint-and-validate on this file." Your AI agent just got 1381+ new capabilities. GitHub Repo: https://lnkd.in/dj49kCAm

This is a major moment for AI coding tools. The Claude Code source map leak has led to over 1,000 GitHub clones. This is good for the industry - it creates more competition for Anthropic and OpenAI. Teams building custom coding agents will benefit most. You can now adapt these features to improve your tools, get better results, and lower costs. I will check back in a month to see the impact on the broader industry. I hope this pushes Anthropic to innovate faster. Read this for tips on improving your Claude Code workflow (by MindStudio) https://lnkd.in/eYA5vJUj

AI agents are amazing until they forget your repository's architecture 10 messages into the chat. I got tired of paying for massive API token contexts just to remind Cursor and Claude how my database migrations work every single session. Semantic search (RAG) wasn't cutting it—when my code changed, the AI would hallucinate outdated context. So, I built Memographix. It’s an open-source, local memory layer for AI agents that uses the Model Context Protocol (MCP). How it works: - Task Capsules: The AI writes a compressed summary of how it solved a specific problem in your repo. - Staleness Tracking: If you (or another dev) modify the underlying files, Memographix flags the memory as stale so the AI doesn't use outdated facts. - Portability: Works seamlessly across Claude Desktop, Cursor, Copilot, Aider, and Windsurf. I even pinned it against a Kubernetes benchmark to prove the deterministic quality score. Stop dumping 20 files into your prompt just to get a feature shipped. 🐙 GitHub (Code & Benchmarks): https://lnkd.in/dQqG2SCy 📦 PyPI: pipx install memographix Give it a spin and let me know if it saves you as many API tokens as it saved me.