The fenbs blogAll articles
Page 18 of 48
- Vibe Coding Prompts: Templates That Keep a Project SaneSix copyable prompts for the moments a vibe-coded project goes wrong: starting it, adding a feature, fixing a bug, reviewing, refactoring and getting ready to ship. Each keeps the change small, asks for a plan first and names the check, and each part is explained. · 8 min read
- Vibe Coding vs Low-Code vs No-CodeAll three promise software without a traditional development project, and all three now use AI. The difference is what you end up holding: source code you maintain, a model on a vendor’s platform, or a configuration that only runs where it was built. Ownership, maintenance, lock-in, security review and the path to production, compared. · 7 min read
- GitHub Copilot Agent Mode Not Showing or Not Working: FixesWhen Agent is missing from the picker, or a session starts and goes nowhere, the cause is usually one of seven things: an old VS Code, the wrong account, a setting or policy, the session target you picked, an extension in the way, AI credits used up, or a different editor. How to tell which, and how to report the rest. · 7 min read
- Cursor Agent Not Working: Fixes for Stuck and Failed RunsA Cursor agent that times out, sits on a terminal command, stops at a usage limit or cannot see your files is usually telling you which of eight things went wrong. How to read the symptom, the fix Cursor documents for each, and what to send support when none of them works. · 7 min read
- Using Gemini CLI in VS Code: The Companion ExtensionGemini CLI runs in VS Code’s terminal, and the Gemini CLI Companion extension connects the two: the CLI sees your open files and selection, and its edits open in VS Code’s diff viewer. How to install and enable it, what it shares, which editors it supports, and how to fix a connection that will not start. · 7 min read
- Power BI MCP Server: What an AI Assistant Can Do With Your ModelsMicrosoft now documents three MCP servers for Power BI: one that edits semantic models, hosted or local, and two that query them. What each can do, how it signs in, its status, how to add it to VS Code, Claude or ChatGPT, and what never to let it change. · 8 min read
- Linear in ChatGPT: The App and What It Can DoLinear reaches ChatGPT two ways: the Linear plugin in ChatGPT’s directory, and Linear’s own MCP server added as a developer-mode app. What each is, the steps, what ChatGPT can read and change, when it asks first, and the limits. · 7 min read
- AI Financial Management: Budgeting, Reports, and ReviewUse AI to organize budget assumptions and draft financial commentary while people verify the sources, calculations, and decisions. Keep the reporting rules and open questions available for the next review. · 5 min read
- AI Governance Consultant: Scope the Work and Choose OneChoose an AI governance consultant by the decisions, evidence, and handoff your team needs. Define the engagement before comparing credentials or accepting a generic policy package. · 8 min read
- AI in Court: A Verification Workflow for Legal TeamsUse AI to organize litigation work while lawyers verify the authorities, the record, and the instructions governing each filing. Keep the checks specific enough that another reviewer can repeat them. · 5 min read
- AI Notes Keeper: Choose Where Your Work Should LiveAn AI notes keeper should help you save, find, and maintain useful information. Compare a note archive, a searchable knowledge tool, and a shared record of work. · 8 min read
- Claude Knowledge Base: Build a Current Project ReferenceBuild a focused reference set for Claude, distinguish source material from instructions, and test answers against documents you know. Keep the current decisions visible as the project changes. · 6 min read
- Claude Projects: Organize Files, Instructions, and Ongoing WorkA Claude project groups related work around reusable files and instructions. Learn what belongs in one, how to start, and what to preserve when the work moves beyond Claude. · 5 min read
- Claude Projects Not Syncing Across Devices: What to CheckCheck the account, organization, project location, and connection before assuming a Claude project is lost. Preserve the latest work, then narrow down what is missing on each device. · 5 min read
- Claude Prompts for Product Managers: Evidence Before PlansUse Claude to organize discovery evidence, challenge assumptions, compare options, and review scope. These prompts keep unsupported claims and product decisions visible. · 5 min read
- Context Window on Bedrock: Budget a Real RequestAmazon Bedrock has model-specific context limits. Identify the model and request path, count the complete input, and leave room for output and later tool results. · 5 min read
- GAO AI Accountability Framework: A Practical Work GuideUse the GAO framework as a reference for AI accountability. Turn review questions into practical work with named owners, evidence, and human decisions. · 5 min read
- Gemini Deep Research: From a Question to Checked FindingsGive Gemini Deep Research a bounded question, review its plan, and check the evidence behind its report. Keep accepted findings separate from open questions. · 5 min read
- Notion AI Privacy: Check Access, Retention, and SharingNotion AI privacy depends on the content involved, who can reach it, and the terms for the features you use. Here is a practical review you can turn into clear team rules. · 5 min read
- A Kanban Board for Claude Code: Plan the Work, Let Claude Pull the Next CardPeople decide what is ready and in what order; Claude Code takes the top card, moves it, comments as it goes, and hands it back for review. How to shape the lanes, the cards and the limits so that works. · 7 min read
- Claude Code Tasks vs To-dos: What Each Is For and Where They LiveClaude Code’s to-do list and its task list are the same idea in two generations: a checklist Claude keeps for itself while it works. What each tool does, what survives the session, and where a shared board takes over. · 6 min read
- Claude Code Task Tool vs Subagents: Which One Runs Your Work?Claude Code’s Task tool and its subagents are not rivals: one dispatches, the other does the work. How subagents are defined, how Claude hands work to them, what they can and cannot see, and how to keep track of what each one did. · 6 min read
- An AI agent task board: what it needs that a normal board does notA board built for people assumes someone who reads the screen, remembers yesterday and hesitates before a big change. An agent does none of that, so the board has to do it instead. Eight things it needs. · 7 min read
- An issue tracker for AI agents: filing, triage and fixAgents now do both ends of an issue: they notice problems while working and they fix the ones they are given. That changes how an issue should be written, who decides what happens next, and why searching before filing stops being optional. · 6 min read
- Kanban for AI agents: how lanes change when agents do the workThe kanban practices still apply when some of the workers are agents. What changes is where you set the work-in-progress limit, how precisely each lane’s exit rule is written, and where the queue forms. · 6 min read
- How to audit AI agents in a small teamA record of what your AI agents did is only useful if somebody reads it against what they were meant to do. Here is a monthly audit a team of two to twenty can run in under an hour. · 8 min read
- MCP security best practices for teamsNine practices that keep a team safe when AI assistants connect to its tools over MCP, drawn from the MCP specification and OWASP guidance, and what each looks like in practice. · 8 min read
- MCP security risks: what can go wrong when an AI connects to your toolsEight ways an AI assistant connected over MCP can be turned against the people it works for, each from a primary or reputable source, and the one mitigation that matters most for each. · 8 min read
- Human in the loop for AI agents: where people should step inApproving every step an agent takes is as useless as approving none. A person belongs at three points: before anything hard to undo, when work is called finished, and when the job itself changes. Everything else can run. · 7 min read
- How to Use AI Agents in Project Management, Step by StepA seven-step plan for a team lead: choose the jobs agents do well, connect one assistant with narrow access, write cards it can finish, set a review rule, run a two-week pilot, measure it, and only then widen it. · 7 min read
- Building a Project Manager Subagent in Claude CodeA worked subagent definition for Claude Code that triages your board, writes plans on cards and files what it notices, with a tool list and a hook that stop it from closing work. · 7 min read
- Running Claude Code Tasks in Parallel Without Losing TrackWorktrees stop parallel Claude Code sessions from editing the same files. Nothing in Claude Code stops two of them picking up the same job. How to run several at once, and how one board keeps them from colliding. · 7 min read
- Claude Code Background Tasks: How to See, Stop and Track ThemShells, subagents, monitors and whole sessions can all run in the background in Claude Code. How to start each kind, where to see it, how to stop it, and how to keep a record that outlives the list. · 6 min read
- How to Stop a Claude Code Task Midway (and Keep a Record)Press Esc. Then decide whether to keep what Claude did, rewind it, or hand it on, and write down where it stopped so the next session does not start from nothing. · 6 min read
- Claude Code Scheduled Tasks: Setting Them Up and Keeping OversightClaude Code can run on a schedule inside a session, on your desktop, in the cloud, or from cron and CI. Nobody is there to approve anything when it does, so the permissions are decided in advance and the results have to land where a person will read them. · 7 min read
- Using Claude Code for Non-Coding Tasks: A Practical GuideClaude Code reads files, edits documents, runs scripts and searches the web, which makes it useful well outside a codebase. Five real non-coding jobs, the permissions each one needs, and how to keep a record of what it did. · 7 min read
- Managing Multiple Projects in Claude Code With CLAUDE.mdSeveral repositories, or one monorepo with many packages: which CLAUDE.md files Claude Code loads depends on where you start it. How the layers stack, what imports and --add-dir really do, and how to point each project at its own place on one board. · 7 min read
- Claude Code Project Structure: Best Practices for Real ReposWhat goes in CLAUDE.md, what goes in the .claude/ folder, which MCP servers belong in the repository and which belong to you, what stays out of git, and why your list of work should not be a file in the repo. · 6 min read
- Claude Code Tasks vs Beads vs a Shared Board: Tracking Agent WorkThree ways to keep track of what a coding agent is doing: Claude Code’s built-in task list, Beads, and a hosted board that people and assistants share. What each one is, where it lives, and who it is for. · 7 min read
- Claude Code Tasks vs the Ralph Loop: How Each Keeps an Agent GoingClaude Code’s task list keeps Claude oriented inside a session. The Ralph loop keeps it working by feeding the same prompt back until the job is done. What each one is, how they differ, and where people and a board come in. · 7 min read
- Taskmaster and Claude Code: How Task Files Drive an AgentTaskmaster turns a product requirements document into a dependency-ordered task file that Claude Code works through, over MCP or from the command line. How it works, where its tasks live, and how it sits next to a shared board. · 6 min read
- Vibe Kanban and Claude Code: What a Kanban View Adds to Agent WorkVibe Kanban puts coding agents such as Claude Code on a kanban board, each in its own git worktree, with the diff waiting for review. What it is, how it runs agents, and how that differs from a shared board where people and assistants are members. · 7 min read
- Claude Code Task Stuck? Common Causes and How to RecoverWhen Claude Code stops making progress it is nearly always waiting on something: you, a shell command, a background job, a full context window, an MCP server or a usage limit. How to tell which, and how to get it moving again. · 8 min read
- Claude Code for Product Managers: A Board-First WorkflowYou do not need to write code to get real work out of Claude Code. Keep the backlog on a board, write cards an assistant can pick up, and use Claude Code to draft specs, sort feedback, build throwaway prototypes and read what the engineers’ assistants did. · 7 min read
- Claude Code Plan Mode: Plan First, Then Track the WorkPlan mode makes Claude Code research and propose before it edits anything. How to enter and leave it, what it does and does not block, and how to turn an approved plan into cards so the plan outlives the session. · 7 min read
- AI Task Managers vs Task Boards With AI MembersAn AI task manager usually means an app that decides when you do things. A task board with AI members is a place where an assistant does some of the things. They answer different questions, and most people need only one of them. · 7 min read
- A Shared AI To-do List for You and Your AssistantNot an app that nags you, and not a list trapped inside a chat. One to-do list that you and your AI assistant both read and change: it adds what you mention, you tick things off, and it tells you where you stand. · 7 min read
- AI Agent Task Examples: 20 Tasks You Can Hand OverTwenty real cards an AI assistant can take from a board and finish, across engineering, product, support, operations and personal work — each with what done means and whether a person should check it first. · 7 min read