You don't need to learn the vocabulary of computers.
They should learn yours.
Claude is smart. But ask "why is my app slow?" and you get:
"The bottleneck appears to be N+1 query generation in the ORM layer, causing redundant round-trips to the database."
You close the tab knowing less than before.
non-dev-output steps in before Claude types a single word. Automatically.
Scenario 1 — You didn't follow the explanation:
You: "I still don't get what a database index is"
Claude: ──────────────────────────────────────────────────────────
By analogy ↓
A dictionary has A–Z tabs on the side.
To find "serendipity" you flip to S — not page one.
A database index is that tab.
A bit more space on disk, but finding rows gets fast.
In reality ↓
CREATE INDEX idx ON table(column);
Use EXPLAIN to confirm the index is being hit.
Only index columns you read often — indexes slow down writes.
No command typed. The plugin caught "I still don't get" and switched modes.
Scenario 2 — You asked for a blog post:
You: "Write a blog post about why coffee makes you more productive"
Without plugin:
Coffee has many effects on productivity.
Firstly, caffeine improves focus.
Secondly, it reduces fatigue.
In conclusion, coffee has a positive impact on productivity.
With plugin:
Every morning, one cup. Just to wake up.
But caffeine does something more specific than that —
it blocks the signal that tells your brain it's tired.
While that signal is paused, your thinking gets sharper.
We call that focus.
No command. The plugin detected a writing request and cut the boilerplate.
claude plugin marketplace add https://github.com/calmtiger86/non-dev-output
claude plugin install non-dev-output@non-dev-outputRestart Claude Code. Done.
omc install https://github.com/calmtiger86/non-dev-outputRestart Claude Code. Done.
Step 1 — Make sure Node.js is installed (v18 or higher).
Run node --version in a terminal. If a version number appears, you're ready.
Step 2 — Clone this repository:
git clone https://github.com/calmtiger86/non-dev-output
cd non-dev-outputStep 3 — Run the installer:
node install.jsStep 4 — Restart Claude Code. Done.
No npm install. No config files. No API keys.
The plugin watches for these — no commands needed.
When you're confused:
| What you type | Language |
|---|---|
| "I don't understand", "what does that mean", "explain simply" | English |
| "이해가 안 돼", "무슨 말이야", "쉽게 설명해줘" | Korean |
| "什么意思", "看不懂", "简单说一下" | Chinese |
| "too technical", "what?", "explain it again" | English |
When you're writing:
| What you type | Language |
|---|---|
| "write a blog post / newsletter / caption" | English |
| "블로그 써줘", "카드뉴스 만들어줘", "뉴스레터 작성해줘" | Korean |
| "写博客", "写公众号文章", "写小红书" | Chinese |
Every session start
└── Loads the "two-block rule" into Claude's context
(analogy block + reality block, no mixing)
Every message you send
└── Scans for confusion or writing signals
├── Confusion detected → injects [CLARITY-MISS] instruction
└── Writing detected → injects [WRITING-INTENT] instruction
Key design: no false positives. Writing mode needs both a subject (blog, newsletter…) and an action (write, draft…) in the same message. Casually mentioning a blog won't set it off.
non-dev-output/
├── hooks/
│ ├── hooks.json ← auto-registers on install
│ ├── session-start.mjs ← loads rules at session start
│ └── clarity-keywords.mjs ← pattern detection (self-contained)
├── rules/
│ └── output-clarity.md ← the two-block rule
├── skills/
│ ├── explain-by-analogy/ ← extended analogy skill
│ └── ko-humanche-calmta/ ← Korean prose correction skill
├── install.js ← cross-platform installer
├── install.sh ← Mac/Linux shortcut
└── install.ps1 ← Windows shortcut
# Remove plugin files
rm -rf ~/.claude/plugins/non-dev-output
# Open ~/.claude/settings.json and delete the two entries
# under "hooks" that reference "non-dev-output".MIT © calmtiger86