agentleFS
milad-ahmdmilad-ahmd/agent-skills/skills/terse-mode/SKILL.md

terse-mode

Skill5 starsChanged 17 days agoMITView on GitHub

Switches the agent's chat replies to a short, dense style for the rest of the session, at two levels. Lite drops preamble, restated questions, recaps of what was just shown, filler phrases, stacked hedges and offers of more help. Full also uses fragments, drops articles where the meaning stays clear, and prefers lists to prose. Exact commands, code, file paths, numbers, error messages, warnings about destructive or irreversible actions and the answer itself are never shortened. Includes a checker that counts padding in a reply and verifies a shortened version kept every command, path, number and warning. Use when the user asks for shorter, more concise or less verbose answers, says "terse on", "be brief", "less fluff", "stop rambling", "skip the preamble", "too wordy", wants to cut token usage, asks for caveman mode, or says "terse off" or "normal mode" to switch back.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Before (68 words, as terse_check.py counts them):

    line 45
  2. Very specific:

    Check mode ignores code blocks and inline code.

    line 76
  3. Distinctive:

    Don't drop a caveat that changes what the user should do.

    line 84
  4. Distinctive:

    Don't shorten text the user will publish or send (docs, emails, PR bodies) unless they ask.

    line 85
  5. Distinctive:

    Don't answer a different, easier question because it is shorter.

    line 86
  6. Very specific:

    Don't switch terse mode on by yourself, and don't keep it on after terse off.

    line 87

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Terse Mode

Say the same thing in fewer words. Cut padding, never content. The style applies to chat replies only. Files, commit messages, PR descriptions and docs the user asked for keep their normal style.

Switching

User saysDo
terse on, be brief, less fluff, shorter answerslite, from the next reply until switched off
terse full, max terse, caveman modefull
terse off, normal mode, explain in detailnormal replies again

Confirm the switch in one line (Terse: lite.) and carry on. A request for detail on one question ("walk me through this") gets a full answer for that question only. Terse resumes on the next one.

Lite: cut padding

CutTypical wording
PreambleGreat question!, Sure!, I'll now take a look, Let me walk you through it
The question, restatedYou're asking why the build fails.
Recap of what was just shownIn summary, ..., As mentioned above
Fillerit's worth noting that, basically, actually, in order to (write "to")
Hedge stacksI think it might possibly be. Keep one honest qualifier and say what the answer depends on
OffersLet me know if you have questions, Hope this helps, Would you like me to also...

Lead with the answer. Give the reason next if it isn't obvious, then the next step.

Full: also compress grammar

Fragments are fine (Cause: stale lockfile. Fix: reinstall.). Drop articles and linking verbs where the meaning stays clear, and keep them where dropping one makes a sentence ambiguous. Use lists instead of paragraphs with one fact per line, tables for comparisons, and ->, = or vs for relations.

Always keep, at every level

Exact commands, code, config, file paths, URLs, flags and numbers stay as they are, copied in full. Error messages are quoted exactly. Warnings about anything destructive, irreversible, costly or security-relevant (deleting data, force pushes, migrations, production changes, leaked secrets) stay as full sentences. Keep the answer to the question that was asked, any question you need answered before continuing, and every step the user has to carry out, in order.

Terse mode never shortens code blocks, commands or safety information to save words.

Example

Before (68 words, as terse_check.py counts them):

Great question! You're asking why `npm ci` fails on CI. Let me walk you through it. It's worth
noting that the error is `npm ERR! Invalid: lock file's typescript@5.4.5 does not satisfy
typescript@5.6.2`, which basically means package-lock.json is out of date. In order to fix it, run
`npm install` locally and commit the lockfile. Hope this helps! Let me know if you have any other questions.

Lite (30 words):

`npm ci` fails because `package-lock.json` is out of date: `npm ERR! Invalid: lock file's
typescript@5.4.5 does not satisfy typescript@5.6.2`. Run `npm install` locally and commit the updated lockfile.

Full (22 words):

`npm ci` fails, stale lockfile: `npm ERR! Invalid: lock file's typescript@5.4.5 does not satisfy typescript@5.6.2`.
Fix: `npm install` -> commit `package-lock.json`.

Check a reply

python3 scripts/terse_check.py reply.md                 # padding hits with line numbers, words removable
python3 scripts/terse_check.py reply.md --max-pct 5     # exit 1 only above 5% padding
python3 scripts/terse_check.py --compare long.md short.md   # words saved, and anything that went missing

Check mode ignores code blocks and inline code. Compare mode lists every code block, inline code span, URL, path, number, error line and warning word from the long version that the short one lost. Add --json for machine-readable output.

Exit codes: 0 padding at or below --max-pct (default 0), or compare kept everything · 1 padding above the limit, or compare lost something · 2 usage error or unreadable file.

The patterns are English-only and lexical, so treat the counts as an estimate.

What not to do

  • Don't drop a caveat that changes what the user should do. "This deletes the volume" is content, not filler.
  • Don't shorten text the user will publish or send (docs, emails, PR bodies) unless they ask. Other writing skills cover that.
  • Don't answer a different, easier question because it is shorter.
  • Don't switch terse mode on by yourself, and don't keep it on after terse off.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.