Practical Claude Code guides: CLAUDE.md, skills, hooks, subagents, MCP, permissions, cost and the workflow on a real project.
Compare Claude Code, Cursor, and Copilot by paradigm and task: autocomplete and suggestions in the editor versus an autonomous agent in the terminal, with a hands-on combo and review always.
Set up a short CLAUDE.md and safe permissions in Claude Code: plan permission mode, a ready allow, ask, and deny list, a template with a test command, and a weekly review with /permissions.
A 30-minute script for your first steps in Claude Code: practice branch, reading before editing, minimal change, review with git diff and npm test, first commit, and one-shot with claude -p.
Step by step for installing Claude Code on Windows, Mac, and Linux, signing in with the right account, understanding pricing, and confirming with claude doctor.
A welcoming guide to installing, configuring, and using Claude Code safely, even with zero prior experience.
The list of git operations auto mode refuses on its own, worktrees for parallel work, and why /rewind does not undo what a background review wrote.
The documentation says a plugin runs arbitrary code with your privileges. The five-minute checklist using the tools already shipping in Claude Code.
One of the three stopped being a separate category. The four questions that pick the right mechanism, and the precedence table that flips order between them.
48,000 real turns measured across 67 sessions. Where the money goes, why turn 400 costs three times turn 10, and how much compaction gives back.
Eighteen documented behaviors that degrade in silence. The list is organized by symptom: what you see before you know what caused it.
Three mechanisms make Claude Code look forgetful, and each has its own fix. What compaction keeps, what it drops, and what costs you.
Deny beats ask, which beats allow, and specificity changes nothing. Claude Code's three permission layers and the rules that look protective without protecting.
All three adopted the same open standards and the feature comparison ended in a tie. What still differs is where context lives and who enforces restrictions.
A custom command became a skill. What changes: positional arguments, context injected before the agent sees it, and stacking up to six in one message.
The three installation scopes, the precedence that does not merge fields, and the trap of MCP's local scope, which differs from settings' local.
A subagent solves verbose context, not speed. And its precedence is the reverse of a skill's — which catches anyone running both under the same name.
A hook that denies blocks the tool even under bypassPermissions. Hooks tighten restrictions and never loosen them, and that is what makes them the policy layer.
A skill does not unload after use: its body enters the conversation and stays for the whole session. That changes how you write one and how many to install.
Two serious studies of coding agents reached opposite results. The difference lies in the five layers you build around the agent.
CLAUDE.md is context delivered as a user message, not applied configuration. Understanding that changes what you write in it and what has to become a hook.
