Same work. Same precision. Words you already know.
Always on · No invoking · Code stays exact
Other tools simplify a topic when you ask them to. This one changes the voice your agent works in, from the first word of the session. One install, then nothing to remember.
You know the reply. It's correct, it's thorough, and you read it three times. Somewhere in there is the one sentence you needed, wearing four words you'd have to look up.
Layman's Terms makes your agent say that sentence first, in words you already own.
You ask what a browser extension does with your login.
The extension proxies your existing browser session, so the credential never leaves Chrome. Doctor deliberately refuses to make network calls, so it can only report configured, never working.
With Layman's Terms:
It borrows the logins already in your Chrome — your login details never leave the browser. The
doctorcheck only reads your settings — it never tests the real connection — so it can say 'set up' but not 'working'.
Same facts. Same doctor, unrenamed. One of them you can act on immediately.
| Rule | What it does |
|---|---|
| Lead with meaning | The first sentence says what happened and what it means for you. Detail comes after. |
| Define or replace jargon | Every term of art gets six plain words of explanation the first time, or gets replaced. |
| Numbers over adjectives | "3 of 15 work", not "several work". "Took 40 seconds", not "took a while". |
| Three-part status | What changed → what works now → what you must do. In that order. |
| Code stays exact | Commands, paths, error text and real API names are never simplified or renamed. You may need to type them. |
| No condescension | Plain is not childish. No "in simple terms…", no re-explaining what you just said. |
| Short sentences | One idea each. If it needs a comma map to navigate, it gets split. |
/plugin marketplace add sobirjoon/laymans-terms
/plugin install laymans-terms@laymans-terms
(Two separate prompts — send the first, then the second.)
Restart your session. Checking layman's terms mode... at startup means it's working; /laymans-terms:status confirms it any time.
macOS, Linux and Windows. On Windows it uses Git Bash, which Claude Code already requires. No Node, no network calls, no config file.
| Command | What it does |
|---|---|
/laymans-terms:plan |
Explains the plan you're about to approve: what changes, what doesn't, the risks, the decision being asked of you. |
/laymans-terms:codebase |
Tours the repo you're standing in: what it is, the parts that matter, one action traced end to end, where to start reading. |
/laymans-terms:changes |
Explains what changed. Uncommitted work by default, or name a commit, branch or PR. |
/laymans-terms:error |
Decodes an error: what broke, how bad, likely cause, next step. |
/laymans-terms:on · :off · :status |
Turn the voice on or off, or ask which it is. Turning it on also recaps your last reply in plain words. |
/laymans-terms:help |
Quick reference for the commands above. |
Off stays off. :off writes a flag file at ~/.claude/.laymans-terms-off and the startup script checks for it, so the setting survives compaction, resume, and every session after this one — until :on removes it. While off, the plugin injects nothing at all, so it costs you nothing.
Plain chat works too: stop laymans terms or technical mode turns it off for the rest of the conversation, laymans terms turns it back on. Chat only affects the current conversation; the commands are what persist.
Uninstall with /plugin uninstall laymans-terms@laymans-terms.
One shell script and eight small command files. That's the whole plugin.
.claude-plugin/plugin.json # plugin identity
.claude-plugin/marketplace.json # lets /plugin marketplace add find it
hooks/hooks.json # runs the script at session start
hooks/session-start.sh # prints the rules, unless the off flag is set
commands/*.md # the 8 slash commands
tests/off-switch.sh # checks the flag works both ways
session-start.sh prints the seven rules as JSON, and Claude Code hands them to the model as session context. No network calls, no dependencies, nothing to configure. It's the same mechanism Anthropic's own explanatory-output-style plugin uses.
Verify a checkout with bash tests/off-switch.sh. It runs against a throwaway config directory, so it can't disturb your own setting.
What does it cost me? About 2 KB of instructions per session, roughly 500 tokens. One short follow-up question costs about the same. If that's still too much, use an on-demand skill instead and pay the cost of remembering to invoke it.
Will it dumb down my code?
No. It governs prose only. Rule 5 is a hard boundary: code, commands, file paths, error text and real API names pass through untouched. retrieveRawInitData never becomes "the data getter".
How is it different from the "explain simply" skills? Those are tools you reach for. This is a voice you install. The difference shows up on the replies where it wouldn't have occurred to you to ask.
Does it need a config file? No.
MIT