terse plugin
The plugin sets a reply shape for Claude Code: one sentence with the answer, a list or table, one caveat, then stop.
- 54 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The rules change reply style, and the meter command edits the `statusLine` setting.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/lowenbjer-claude-terseDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The rules change reply style, and the meter command edits the `statusLine` setting.
This is third-party code. Review the repository files before installing.
What it does
terse removes the essay habit from Claude Code replies: slogans, metaphors, a recap of what was just done and contrast turns of phrase. The rules run 877 words: 22 rules and 18 before and after pairs. The reply shape is one sentence with the answer, then a list or table, at most one caveat, then stop. Each sentence names its topic. The rules ship as an output style and apply while the plugin is enabled, and a `SubagentStart` hook hands the same text to subagents. The `/terse:install-meter` command copies a 26-line context meter script once and adds a `statusLine` entry to settings. The plugin makes no network calls and collects nothing.
Who it is for. Claude Code users annoyed by long replies with slogans and recaps who want answers to the point.
Good fit when
- Every question comes back as an essay but you want a short answer with a caveat.
- A subagent reply sounds different from the main agent and one style is needed.
- You want to watch context fill through an indicator in the status line.
- You measure how much shorter replies become on your own prompt set.
Not a fit when
- You want an agent's extended reasoning, since the rules cut volume on purpose.
- You want to save money, since cost falls less than word count.
- Your main agent is not Opus 5.5 or Fable 5.1, with no measurements for other models.
Example request
Explain why the number of database connections keeps growing and answer under the terse rules: an answer sentence, a list, one caveat.Limitations
Measurements were made on Opus 5.5 and Fable 5.1 as main agents, with Opus 5 and Sonnet 5 data only as subagents. The 150-word cap shortens replies but does not guarantee the limit, as long analysis questions still run over. Output tokens fall 42 to 55 percent while total cost falls 2 to 12 percent, because context makes up most of the cost. Subagents follow their own model, and in the measurements an Explore subagent on Opus 5 kept em dashes. The reminder adds about 100 uncached tokens per call.
How to disable. For a GitHub install, remove the plugin with `/plugin` commands. If you installed the meter, remove the `statusLine` entry from `~/.claude/settings.json`; uninstalling deletes the plugin data directory itself.
FAQ
How do I install it?
From the Claude plugin directory press "Add to Claude Code" or run `/plugin marketplace add lowenbjer/claude-terse` and `/plugin install terse@terse`.
When do the rules apply?
In every new session, after `/clear`, and in an existing session after you exit and `claude --resume`.
Does it collect data?
No. The plugin makes no network calls, hooks read a text file inside the plugin directory and the meter script writes one entry to settings.
How much shorter are replies?
On 20 prompts against a real codebase, word count fell 46 percent on Fable 5.1 and 54 percent on Opus 5.5.
How do I add my model?
The `docs/models` catalog has one file per model. Run the 12 public prompts with and without the plugin through `bench/run.py` and send results as a pull request.
Discussion
No questions or reviews yet.
Related
Pluginshush quiets running narration in Claude Code and puts the result first: what changed and what comes next.
Two plugins change how Claude Code output looks: replies in the glamour dark style and tool calls as compact one-liners.
A plugin that renders Claude Code answers as a typeset picture via Typst: tables, math, charts and diagrams in the terminal.
A plugin that measures language complexity in coding-agent replies and sends an overly abstract reply back for a rewrite
pordee (พอดี)
pordee
A Claude Code plugin that makes the agent answer in Thai briefly: no polite particles, hedging or greetings
A plugin with output styles for coding agents: short, human answers instead of verbose walls of text
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| terse (this card) | no | 54 | |
| Terse | no | 78 | difference: a card for another `terse` implementation, while this plugin builds on an output style with 22 rules and a subagent hook |
| Caveman | no | 119,585 | difference: `caveman` compresses speech, while terse sets a reply shape and strips slogans and metaphors |
README in short
The README tells why the author built the plugin, how it works, what it gives, its measured limits and how to add your model. Measurements come with models and dates.