Ivan Mišić product · tech · ai

The Toolshed

Skills, plugins, rules, and prompts that survived real work. Pick a shelf, scan the promise, and take what helps.

Skills

Reusable instructions for specific tasks in Claude Code, such as reviewing code or analyzing Search Console data.

All 10 skills

Plugins

Skills and supporting tools bundled into one Claude Code installation.

All 6 plugins

Rules

Project instructions for Claude Code, with packs for shared conventions and specific programming languages.

All 5 rules

Prompts

Ready-to-use prompts for a specific job. Copy, paste, adjust the details, run.

All 46 prompts
A checkout countdown has to earn its delivery promise. A checkout countdown can create urgency or support a real delivery promise. That promise needs clear rules for what changes at zero, which date to show, and how it survives payment. from an articleCopy it → A safe way to edit database content with AI coding tools AI coding tools edit files. Your content lives in a database. This is the pull, diff, push workflow that connects the two without hand-writing SQL against production. from an articleCopy it → a-make-prompt Turn a rough request into a prompt with a defined output, relevant context and checks for completion. mine · Jun 2026Copy it → AdGuard Home: network-wide ad blocking you control Part two of the AdGuard series. Running your own DNS server is better than trusting Cloudflare's. See what every device on your network is doing. Decide what it gets to see. from an articleCopy it → AI isn't coming for your team. It's coming for their busywork. AI's immediate value is removing repetitive work from experienced people. Leaders should start with one workflow, not a headcount target. from an articleCopy it → AI made building easy. Knowing what to build just got expensive. AI can make execution cheaper, but it also exposes weak decisions faster. The valuable skill is defining work clearly enough to test whether it matters. from an articleCopy it → Block ads on every device in 5 minutes with AdGuard DNS A simple DNS change blocks most ads on your phone, TV, and every other device. No apps to install, no technical skill required. Takes about five minutes. from an articleCopy it → Building the rest of your site Use the home page as a reference, then build and review each remaining page one at a time. from an articleCopy it → Building your first page Build the home page first, review it in the browser, then record the decisions that should carry into every other page. from an articleCopy it → Capturing email and calendar without a mailbox API Microsoft Graph and IMAP were off the table, so I used Power Automate, OneDrive sync, and one 15-minute scheduled task to pull email and calendar into a local inbox. from an articleCopy it → Claude Code and the context window on bigger projects My website is close to a thousand files. I keep Claude Code useful by giving each session one task, naming the files involved, and keeping the project instructions current. from an articleCopy it → Claude Code permissions: what to allow and what to block Claude Code can change real files and run commands. Check the permission mode, protect secrets, review the diff, test the result, and keep a restore path. from an articleCopy it → Codex vs Claude Code: two agents on one codebase I run Claude Code and Codex daily, sometimes on the same repository. My session logs show the context cost, while the working tree shows the coordination cost. from an articleCopy it → Creating reusable Claude Code skills If you keep pasting the same prompt, save it as a small Claude Code skill. You can run it with a short command and add templates or scripts later. from an articleCopy it → Don't replace your billing system (build around it instead) The graveyard of failed digital transformations is full of billing system replacement projects. The smartest companies build abstraction layers instead. from an articleCopy it → Fixing bugs with Claude Code (without losing your mind) Something broke. Describe what you expected, show what happened, let Claude investigate, then test the proposed fix. from an articleCopy it → Four calendar days, one person, one full redesign A full site redesign in four calendar days, from design system to admin stack. The real question is what product teams should do with the time AI gives back. from an articleCopy it → Giving Claude Code better instructions Give Claude Code the problem, relevant context, and the result you expect. A specific instruction reduces guessing, but you still need to review the result. from an articleCopy it → How Claude Code sees your files The folder where you start Claude Code is its main working directory. Keep one project in scope, then add other directories only when the task needs them. from an articleCopy it → How to undo Claude Code changes with rewind and Git Claude Code will sometimes make the wrong change. Inspect the diff, use rewind for recent mistakes, and keep Git for a durable restore point. from an articleCopy it → How to use Claude Code subagents for parallel work Use Claude Code subagents when a focused task can stand alone. Split independent checks, ask for one combined result, and keep dependent work with one owner. from an articleCopy it → Installing Claude Code without the old Node.js setup Install Claude Code with Anthropic's current native installer on Windows, macOS, or Linux. No Node.js, no npm, and the fixes for when the command is not found. from an articleCopy it → Keeping your Claude Code project organized Give each file an obvious home, document the stable rules, and test what your deployment actually includes. from an articleCopy it → Klarna's AI U-turn isn't a failure story Klarna reported faster support and stable satisfaction, while its CEO later said its customer-service cost-cutting went too far. Product leaders need to read both. from an articleCopy it → Making it look good (and fixing what doesn't) On this site, building the structure was the short part. I spent much longer refining the design. from an articleCopy it → Most of my AI pipeline isn't AI By volume, most of my AI pipeline is plain Python. The model is a small, expensive part I call as little as possible. The script-first split, and why it wins. from an articleCopy it → My AI was smart. It just couldn't see my job. My work context was scattered across email, meetings, documents, and notes. So I wired it into a private Obsidian vault Claude Code can read. from an articleCopy it → One app, many markets: the central vs. local paradox Build centrally and local markets hate you. Build locally and you drown in fragmentation. The model that actually works for multi-market digital products. from an articleCopy it → One command a day: the skills that run the vault The whole control layer of my work system is a folder of markdown files. How a real workflow is built from Claude Code skills, rules, and parallel sub-agents. from an articleCopy it → Planning before you build with Claude Code Write down the pages, content, and design constraints before Claude creates files. A short plan gives you something concrete to review and reduces avoidable rebuilding. from an articleCopy it → Saving your work with Git and going live Two problems to solve: keep a recoverable copy of your work and publish the site without exposing private files. from an articleCopy it → Stop trying to digitalize everything The biggest trap empowered digital teams fall into: trying to replace every legacy process. Solve for 80% of cases with clean journeys. The rest? Let it go. from an articleCopy it → Teaching Claude Code your standards with rule files Use CLAUDE.md for project-wide facts and rule files for detailed standards, including rules that load only for matching paths. from an articleCopy it → The app consolidation playbook Most app portfolios are 80% accident, 20% strategy. The interesting bit isn't deciding what to kill. It's keeping what works locally without paying for it everywhere. from an articleCopy it → The Claude Code plugins I built and run I put the Claude Code tools I reuse across projects into one public toolshed: generators, workflow commands, anti-slop checks, search visibility, rule packs, and prompts. from an articleCopy it → The laundry list trap: why half your roadmap doesn't matter We analyzed engagement across an entire app feature set. Over 40% of the scope generated less than 2% of engagement. This is how to avoid building features nobody uses. from an articleCopy it → The permission tax: when growing teams ship less, not more It took a month to build and three to four months to get permission. Team growth can slow delivery when every decision needs another approval. from an articleCopy it → The person logging in is not always the customer. Login proves who is at the screen. It does not prove what they own, may manage, or can buy. Model those questions separately before the edge cases arrive. from an articleCopy it → What is vibe coding (and why you already know how to do it) Andrej Karpathy called it "fully giving in to the vibes." Collins Dictionary made it Word of the Year. What vibe coding actually means, when it works, and how everything you've learned so far has been building toward it. from an articleCopy it → What using Claude Code actually looks like Using Claude Code is a loop: ask, change, check, decide. Here is what a normal session looks like, six commands worth learning first, and how to recover when it goes wrong. from an articleCopy it → Where to go from here You have a working website and enough Claude Code experience to build the next thing. Pick one real problem, define done, and finish the first part. from an articleCopy it → Why renaming IT to product didn't fix anything You announced a digital transformation. You hired a CDO. You renamed the IT team to Product. Six months later, nothing has changed except the org chart. from an articleCopy it → Writing CLAUDE.md: permanent memory for Claude Code Every time you start Claude Code, it forgets who you are. CLAUDE.md is a text file that gives it permanent memory. One file, read automatically at the start of every session. from an articleCopy it → Your app isn't a service channel (it should be your business) Most companies treat their app as a self-service cost-saver. The ones winning treat it as their primary business channel. The evolution from service to engagement, in three phases. from an articleCopy it → Your digital team needs to act like a startup Bean bags and hackathons aren't startup culture. The real difference is authority: who decides what to build, who can say no, and who's allowed to be wrong. from an articleCopy it → Your next team is three people and a pile of agents. I expect AI-assisted teams to become smaller and organize around complete workflows. The model works only when the pod owns the outcome and runs what it builds. from an articleCopy it →

On my bench

Not mine, not rehosted. The AI and product tools actually on my board, with the honest take on each.

Open the board
ChatGPT All-purpose AI assistant for writing, research, coding, and productivity Claude Anthropic's frontier assistant with strong reasoning and long context Claude Code Anthropic's agentic coding assistant for the terminal Codex OpenAI's cloud-based agentic coding environment with frontier model Confluence Team wiki and documentation from Atlassian. Forgejo Self-hosted Git forge with a lightweight UI. Google Workspace Gmail, Calendar, Docs, Drive in one suite. Jira Issue tracking and agile project management. Microsoft Teams Chat, meetings, and collaboration in Microsoft 365. Obsidian Local-first markdown knowledge base. Plaud AI voice recorder that turns meetings into transcripts, summaries, and action items Slack Channel-based messaging for teams. Backlog.md Markdown task tracker that lives in your git repo Canva AI Design platform with Magic Studio AI tools DeepSeek Open-weight Chinese reasoning model with free chat and self-host options Figma Collaborative interface design and prototyping. Gemini Google's advanced AI for search, productivity, and creative tasks Google Analytics Web and product analytics at scale. Google NotebookLM AI research assistant for documents and knowledge synthesis LM Studio Desktop app for running local LLMs with a friendly GUI Microsoft Copilot AI assistant across Microsoft 365 Beautiful.ai AI presentation software with automatic design Bolt.new Browser-based full-stack development with WebContainers running Node.js in-browser Cursor AI-first code editor that understands your entire codebase DALL-E OpenAI's image generator, built into ChatGPT ElevenLabs Realistic AI voice generation Gamma AI-powered presentations and docs GitHub Copilot AI pair programmer for code completion Grammarly AI writing assistant for clear, credible communication Lovable AI app builder with Supabase backend and team collaboration Midjourney Leading AI art generator for stunning visuals Miro Collaborative whiteboard for workshops and strategy. n8n Open-source workflow automation with AI nodes Notion Docs, wikis, and project notes in one workspace. Notion AI AI writing and knowledge assistant Ollama Run large language models locally with a simple CLI Outlook Microsoft's email and calendar client. Perplexity AI-powered answer engine with real-time search Tana Supertag-powered workspace for notes, tasks, and meetings

Everything mine on these shelves comes from my own working setup, shared for learning. Test it and adapt it to your project before relying on it; you run it at your own risk.