Od wersji 2.1.50 nie jest to już konieczne. W Claude Code pojawiła się możliwość skorzystania z wbudowanej opcji --worktree. Wywołanie claude --worktree spowoduje utworzenie nowego worktree o losowej nazwie w lokalizacji ./.claude/worktrees. Jeśli chcemy utworzyć worktree o konkretnej nazwie, możemy podać ją w poleceniu: claude --worktree <worktree_name>. Po zamknięciu sesji Claude automatycznie usuwa utworzone worktree oraz powiązaną gałąź, jeśli nie ma zmian w working directory ani nowych commitów. Jeśli wprowadzono zmiany, Claude zapyta, czy je zachować. Jeśli odrzucimy zmiany, zarówno worktree, jak i powiązana gałąź zostaną usunęte.

Vibe Hacking: Claude Code Can Be Turned Into A Nation-State-Level Attack Tool With No Coding At All

Best Practices Clear triggers: Define specific conditions for activation Focused scope: Each skill should do one thing well Informative prompts: Give Claude clear instructions Error handling: Account for edge cases in prompts Test thoroughly: Verify skills work across scenarios

Variables Available VariableDescription$CHANGED_FILESList of modified files$CURRENT_FILECurrently focused file$PROJECT_ROOTProject root directory$GIT_BRANCHCurrent git branch

Trigger Types TriggerDescriptiononFileChangeFiles matching glob pattern changeonCommandUser invokes a slash commandonGitHookGit operations (commit, push)onScheduleTime-based triggers

2. Create skill.json { "name": "test-runner", "description": "Automatically runs relevant tests when code changes", "triggers": { "onFileChange": ["**/*.ts", "**/*.tsx"], "onCommand": "/test" } }

Skill Structure .claude/skills/ └── my-skill/ ├── skill.json # Skill configuration ├── prompt.md # Instructions for Claude └── scripts/ # Optional helper scripts └── helper.sh

Skills are advanced automation capabilities that Claude can invoke automatically based on context. Skills vs Slash Commands FeatureSlash CommandsSkillsInvocationManual (/command)AutomaticTriggerUser types commandContext-basedComplexitySimple promptsScripts + promptsUse caseRepetitive tasksSmart automation

Test thoroughly: Run commands in different scenarios

Include specific files with curly braces: <!-- .claude/commands/optimize.md --> Analyze {src/utils/helpers.ts} for optimization opportunities.

Include context: Reference relevant files when needed

Use variables: $ARGUMENTS for flexibility

Clear instructions: Be specific about expected output format

Single responsibility: Each command should do one thing well

Using Arguments Commands can accept arguments via $ARGUMENTS: <!-- .claude/commands/explain.md --> Explain $ARGUMENTS in simple terms. Provide: - What it does - Why it's useful - Example usageCopy to clipboard Usage: /explain the useCallback hook

Command Locations # Global commands (available everywhere) ~/.claude/commands/ # Project-specific commands .claude/commands/

I simply could not have built this project as well or as quickly without help. And as other developers have noted, this is the help that's showing up.

Although I read each proposed change, knowing the codebase deeply was much more challenging. When I write a new application myself, I'm building an elaborate house of cards in my head, a gossamer structure of interlinked ideas and goals. It's a story I'm telling myself in code—and ultimately, a story I share with users.

"Human in the loop" is necessary, but the current process itself makes the loop stultifying, and encourages the human to take themselves out of the loop. That process is straight up dangerous. The temptation to let it rip is always there, and I didn't even have a boss pressuring me to ship code.

It was so tempting to press 2: "Yes, and accept all changes for this session." Why wouldn't you? If you're accepting them all manually, what's the harm? What's the harm? harm harm harm harm Yeah, that's how you get got in this process. Once you stop scrutinizing the model's output, the probability something goes off the rails approaches 1.

I hated writing software this way. Forget the output for a moment; the process was excruciating. Most of my time was spent reading proposed code changes and pressing the 1 key to accept the changes, which I almost always did. I was basically Homer's drinking bird.

If it works, I'll have my certificate solution, I thought. If it doesn't, at least I'll know more about the technology and its implications. Well, spoiler alert: it works. It's even, near as I can tell, reasonably secure. But good lord, building this way was miserable, even if it was faster than coding it all myself.

I solved it with a two-tier system - short-term memory.md files and long-term topic indexes.

Context length is the maximum number of tokens that the model has access to in memory. The default context length in Ollama is 4096 tokens. Tasks which require large context like web search, agents, and coding tools should be set to at least 64000 tokens.

Set the environment variables:

Quick setup Copyollama launch claude

Recommended Models qwen3-coder glm-4.7 gpt-oss:20b gpt-oss:120b

Further ReadingI’m not gonna pretend to be an expert here (any more than I’m an expert Obsidian plugin developer :p) but here are some resources that helped me figure out Claude CodeKent writes a lot about how he uses Obsidian with Claude Code.This is an incredible hub of resources for using Claude Code for project management, by someone who also uses Obsidian.This take on Claude Code for non-developers helped solidify my understanding of how it all works; it hallucinates less, for one thing.Eleanor Berger has fantastic tips for working with asynchronous coding agents and is incredibly level-headed about the LLM landscape.This article does a great job of breaking down all the nitty-gritty of how Claude Code works.Damian Player has a step-by-step guide on using Claude Code as a non-technical person that goes into more depth.Here’s a tutorial from a pro that breaks down best practices for using Claude Code, like the importance of planning and thinking things through, and exactly why a good CLAUDE.md file matters.

Little Tips for Claude Code + Obsidian

Terminal Practice with GamesSome folks I’ve talked to are a little intimidated by the terminal. Want to practice in a low-stakes way?

But these days I’m not generally trying to do things faster, I’m trying to do them with less attention. All these searches and tasks run in the background, which means they actually get done. When I had to actively sit there and click through things, half of it never happened because something else more important would come up, or I just didn’t feel like doing grunt work just then.

Setting Claude Code Up in ObsidianI was genuinely surprised at how easy the terminal plugin was to install for Obsidian. In Obsidian, I went to community plugins, searched for “terminal,” and installed the Terminal plugin by polyipseity. Then I clicked the “open terminal” button on the left-hand side. That’s it.There’s a dedicated Claudian plugin (subtly different from the Claudsidian solution people), but the Terminal felt a little higher fidelity to how I’m used to doing things, and a little simpler to understand. Plus, Claudian looks great but honestly I don’t think I can live without plan mode, which the readme says it doesn’t currently support. Plan mode is nice because it asks questions, really thinks things through, and can be trusted not to do dumb destructive things.

If you have been following along with me for years you know I don’t hype things just because people are hyping things. But Claude Code finally has made AI a core part of my processes instead of just a thing I use sometimes as an extra source or bonus spell checker or quicker way to reformat files.

The UI feels so intuitive, like an old-school MUD.

My excitement for local LLMs was very much rekindled. The problem is that the big cloud models got better too—including those open weight models that, while freely available, were far too large (100B+) to run on my laptop.

The reason I think MCP may be a one-year wonder is the stratospheric growth of coding agents. It appears that the best possible tool for any situation is Bash—if your agent can run arbitrary shell commands, it can do anything that can be done by typing commands into a terminal. Since leaning heavily into Claude Code and friends myself I’ve hardly used MCP at all—I’ve found CLI tools like gh and libraries like Playwright to be better alternatives to the GitHub and Playwright MCPs.

It turns out tools like Claude Code and Codex CLI can burn through enormous amounts of tokens once you start setting them more challenging tasks, to the point that $200/month offers a substantial discount.

the trade-off: using an agent without the safety wheels feels like a completely different product. A big benefit of asynchronous coding agents like Claude Code for web and Codex Cloud is that they can run in YOLO mode by default, since there’s no personal computer to damage. I run in YOLO mode all the time, despite being deeply aware of the risks involved. It hasn’t burned me yet... ... and that’s the problem.

As-of December 2nd Anthropic credit Claude Code with $1bn in run-rate revenue!

It helps that terminal commands with obscure syntax like sed and ffmpeg and bash itself are no longer a barrier to entry when an LLM can spit out the right command for you.

Claude Code and friends have conclusively demonstrated that developers will embrace LLMs on the command line, given powerful enough models and the right harness.

The year of coding agents and Claude Code # The most impactful event of 2025 happened in February, with the quiet release of Claude Code. I say quiet because it didn’t even get its own blog post!

In November, Claude Code achieved a significant milestone: just six months after becoming available to the public, it reached $1 billion in run-rate revenue