EDIBLOG · AI TIPS · ONE-MONTH REVIEW
I Use All 5 Claude Code Plugins Every Day — An Honest Breakdown
Hi, I’m Edi. A month ago I’d run Claude Code for a full year with zero plugins. Today I have five installed and I use all of them every day. I’ll lay it all out honestly: what’s installed, when each one fires, and the difference between not having it vs having it. No ads, no sponsorship — just my own measured experience.
01 · MY 5 PLUGINSI use all five daily — in the order I installed them
Bottom line. My first year ran on zero plugins. A month ago (2026-04-26) I started with one — frontend-design — and on May 12 I added four more in a single day. All five have now settled into a daily or near-daily cadence.
Here are the five, ordered by how often I use them, with a one-line note on what situation triggers each.
When it fires. Every time I write a post in my v2 infographic style. It auto-applies the magazine card system, light-tone readability, and my dark-box avoidance rules.
I trigger it inside all five of my post-generation jobs (run-daily · run-devtools-daily · run-weekly · run-ai-intro, etc.).
When it fires. Right after an automation job has an incident, or when I need a new guard rule. I’m running 2 PreToolUse hooks + 1 PostToolUse hook.
My flow: check the current state with /hookify:list → write a new rule with /hookify:hookify → it lands in settings.json automatically.
When it fires. Whenever I build a new skill or tweak an existing one. I carry 8 project skills (add-launch-job · blog-style-guide, etc.).
It handles the SKILL.md frontmatter format · description tuning · even the eval in one pass, so the cost of adding a skill is basically zero compared to hand-rolling it a year ago.
When it fires. Two-way messaging between a Claude Code session ↔ my phone. I use it for interviews and instant job-debugging replies.
To avoid a token clash with my existing yongBlogBot, I spun up a separate bot via BotFather and paired it exclusively to the plugin.
When it fires. Debugging publish jobs · inspecting pages. I keep the actual publish code in lib/ Python untouched and use the MCP browser tools only for debugging.
I don’t rip out the publish job code all at once — I observe for a week → confirm it’s stable → migrate gradually.
02 · BEFORE vs AFTERThe gap between not having them and having them is bigger than I expected
Bottom line. For a year I held the “take it or leave it” view. After a month with them, I realized that the time freed up from hand-editing got filled with actual writing time.
Here’s how I did each task without the plugin, and how it changed with it — all in one table.
My year vs my month · all five compared
/hookify:list shows the state. Rule writing is standardized. Less debugging time.Looking at the table taught me one thing. A plugin isn’t a “new feature” — it’s “automating the manual work I was already doing.” All five took over something I’d been doing by hand for a year.
03 · HOW TO FINDI locked in 4 criteria for finding a good plugin
Bottom line. I learned that if you open the marketplace and install everything that looks cool, your menu turns into a mess within days. So I run a 4-step check before installing anything.
My 4 criteria · the filter before installing
1
Map your own workflow first. Standardizing what you already do often is the highest ROI. For me, that was hookify · skill-creator.
2
Official first, external only after vetting. Anthropic’s official ones are verified. For external ones, you must check for token · credential clashes. When I installed telegram (external), I built a separate bot to avoid a bot-token clash.
3
One-week trial → measure ROI. If you’re not using it, remove it immediately. Leaving things installed clutters the menu and causes command clashes. All five of mine survived past the one-month mark.
4
Don’t rip out operational code like publish jobs. Even after adopting playwright, I left my lib/ Python publish code as-is. Observe for a week to confirm no stability impact, then migrate gradually — that’s the safe path.
My install flow — from opening the marketplace to a one-week trial
# 1. Open the marketplace
$ /plugin
# 2. Browse the official category first
↓ prioritize verified plugins
↓ vet external ones for clashes
# 3. Install one at a time (not several at once)
$ /plugin install hookify@claude-plugins-official
# 4. Use for a week → remove if unused
$ /plugin uninstall <the one you don't use>04 · MY RECOMMENDIf you’re just starting, I’d recommend three, in this order
Bottom line. All five are good, but if you’re just starting, don’t be greedy — look at three first. I’d recommend the order skill-creator → telegram → hookify.
MY RECOMMEND · 3+2
as of 2026-05
The first three touch almost every workflow. The other two are optional.
- #1 · skill-creator
- The first step to turning your workflow into an asset. It’s the fastest path to baking your repetitive work into a skill.
- #2 · telegram
- Session ↔ phone, two-way. Getting answers on your phone from inside a Claude Code session touches daily life more often than you’d think.
- #3 · hookify
- Safety net + guard rules. If you have even one automation job, you’ll need a hook at some point.
/hookify:hookifystandardizes the rule writing. - optional · frontend-design
- Only for people who touch blog · magazine design often. If you’re backend-focused, you can skip it.
- optional · playwright
- Consider it if you do publish automation · web scraping often. Migrate operational code gradually, after a week of observation.
CODAIn the end, the answer is: I use all five every day
A month ago I ran a full year on zero. Now I use all five daily, and I can’t go back. The way I see it, a good plugin isn’t a new feature — it takes over the manual work.
— Edi · after one month of use
Edi’s noteDon’t install them all at once. Use one at a time for a week and drop it the moment the ROI isn’t there — that’s how you keep the menu clean.
A plugin isn’t a new feature — it’s the automation of manual work.
References
Claude Code official
Primary data
- EDIBLOG · one month running five plugins, my own asset (first-hand)
- My own setup · 39 automation jobs · 8 custom skills · 56 lib scripts
Disclaimer: this is my personal operating retro. Results may vary by environment. No ads · no affiliates.
This post is part of an English mirror of a Korean dev experiments blog. Cross-posted for the global developer audience.