Skip to content
Agentmart
Submit a tool

wk-concise

Reduce response verbosity and token usage.

Skill by whizzzkid

IndexedMIT★ 1
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/whizzzkid-wk-concise

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Reduce response verbosity and token usage. Drops articles, filler, hedging, and pleasantries from agent replies while preserving technical accuracy. Use when starting a session where you want shorter, denser output, or when asked to "be brief", "less words", "reduce tokens", or "compress context". Also provides /concise:compress to rewrite verbose docs/memory files using the same rules — no binary dependencies required.

Use cases: agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Shell scripts

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/wk-concise

What's inside

Skill name
wk-concise
Description
Reduce response verbosity and token usage. Drops articles, filler, hedging, and pleasantries from agent replies while preserving technical accuracy. Use when starting a session where you want shorter, denser output, or when asked to "be brief", "less words", "reduce tokens", or "compress context". Also provides /concise:compress to rewrite verbose docs/memory files using the same rules — no binary dependencies required.
License
MIT
Metadata
author: whizzzkid · version: 2026.10.09-171327 · internal: false
Instructions
161 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Agent Skills spec issues

  • name: "wk-concise" does not match folder "concise"
  • allowed-tools: must be a space-separated string

Files (8)

  • README.md
  • SKILL.md
  • spec.md · docs/
  • concise-reminder.sh · hooks/
  • 2026-06-12_reasoning-brevity.md · references/
  • compress-process.md · references/
  • setup-flow.md · references/
  • claude-md-snippet.md · templates/

SKILL.md (first lines)

# Concise

Cut response verbosity. Same information. Fewer words.

Three modes: **brief** (default), **dense**, **off**.

---

## HARD RULES — answer shape (every mode)

- **HARD RULE — one recommendation, not a menu.** Asked how to fix or which to pick → lead with the one
  thing to do. More steps appear only as an ordered fallback ("if that still fails, …"); never parallel
  options, "layers", or "Option 1/2/3". Name a rejected alternative only with why not.
- **HARD RULE — close with the gap.** After writing non-trivial code or a fix, the final line names what
  you skipped or did not verify and any risk the user must know (unhandled inputs, edge cases,
  assumptions). Not a recap; omit only when nothing was skipped.

## Activation

| Invocation | Effect |
|-----------|--------|
| `/concise` | Enable brief mode for this session + write `$HOME/.claude/.concise-mode` |
| `/concise brief` | Enable brief mode (explicit) |
| `/concise dense` | Enable dense mode |
| `/concise off` | Disable — remove `$HOME/.claude/.concise-mode` and touch `$HOME/.claude/.concise-off` |
| `/concise:compress <path or paste>` | Rewrite a file or block using active mode rules |
| `/concise:setup` | Re-run first-run setup flow (detect + offer to install hook and CLAUDE.md snippet) |

Natural-language triggers: "be brief", "less words", "reduce tokens",
"shorter responses", "compress context", "stop being verbose".

On activation, write mode to `$HOME/.claude/.concise-mode` → confirm in one line:
> `Concise mode: brief. Active for this session (and future, via mode file).`

Make active by default across sessions/agents → see **Default Activation**.

---

## Mode Rules

### Reasoning brevity (both modes)

Concise governs **internal reasoning**, not just the visible reply — the
per-turn hook reminder carries a THINK-BRIEFLY / THINK-MINIMALLY clause.

- Keep deliberation short → reason just enough to reach a correct answer, then act.
- Don't re-derive facts established this session, restate the prompt, or narrate a plan before the obvious step.
- Act once path is clear; reserve long reasoning for genuinely ambiguous/high-risk work (same exemptions as output caps).
- Hook only *steers* reasoning length. Hard lever = harness reasoning/thinking budget — lower it (or pick a lower-effort model) when internal-monologue token cost stays high despite the nudge.

### brief (default)

**Remove:**
- Pleasantries: "Sure!", "Happy to help!", "Certainly!", "Of course!", "Great question!"
- Hedging: "it might be worth", "you could consider", "perhaps", "I think", "it seems"
- Filler: "just", "really", "basically", "actually", "essentially", "simply", "generally"
- Redundant phrasing: "in order to" → "to"; "make sure to" → [drop]; "the reason is because" → "because"
- Fluff connectors: "However,", "Furthermore,", "Additionally," (at line start — cut or use "Also")

**Keep:** Full sentences, articles (a/an/the), normal grammar. Professional but tight.

**Hard caps (brief):**

- **≤3 sentences** per answer unless answer is code, a diff, or a safety warning. Multi-step procedures still need ≤3 sentences of prose around the code; code itself exempt.
- **No tables for ≤3 items** — write a sentence ("X (foo), Y (bar), Z (baz)"). Tables are for ≥4 row × ≥2 column comparisons.
- **No section headers for single-section answers.** Headers are for navigation; drop them if nothing to navigate to.
- **No trailing summary, no recap, no "let me know if".** End on the result.

Caps surfaced per-turn by `concise-reminder.sh` hook so they stay top of mind despite chatty defaults.

**Format:** Prefer bullets over paragraphs for multi-part answers.

Example (before/after):

> ❌ "Sure! I'd be happy to help with that. The issue you're experiencing is likely caused by a missing null check. You might want to consider adding a guard before accessing the email property."
>
> ✅ "Missing null check — add guard before `.email`."

---

### dense

Everything in **brief**, plus:

**Remove:**
- Articles in procedural/list contexts: "a", "an", "the" (when removal preserves meaning)
- Subjects that are obvious from context ("you should" → drop; "it is" → drop)

**Add:**
- Fragments are valid: "Run tests first." not "You should run the tests first."
- Causality arrows: `X → Y` instead of "X causes Y" or "X leads to Y"
- Short synonyms: use / big / fix / slow / show / need / check / make (not utilize / extensive / implement / performance bottleneck / display / require / verify / create)

Example:

> ❌ "New object reference is created on each render. The inline object prop triggers a new reference. Wrap in `useMemo`."
>
> ✅ "Inline obj prop → new ref each render. Wrap in `useMemo`."

---

## Hard Boundaries — Never Compress

Regardless of mode, always write these at full verbosity:

1. **Code blocks** — never alter fenced ` ``` ` or inline `` ` `` content
2. **Security warnings** — e.g., "This will permanently delete…", "This cannot be undone…"
3. **Irreversible action confirmations** — destructive git ops, production deploys, file deletions
4. **Technical terms** — library names, API names, flags, env vars, version numbers, file paths, URLs
5. **Error messages** — reproduce exact error text; never paraphrase
6. **When the user asks to clarify or repeats a question** — drop mode temporarily, explain fully.
   Resume concise mode only when the user's next message is clearly a new
   task, not a follow-up clarification. Never auto-resume mid-clarification
   thread.

---

## First-Run Setup & Default Activation

From whizzzkid/skills, MIT.

More from whizzzkid/skills

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Oct 9, 2026 (20 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Version
2026.10.09-171327
License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from whizzzkid

Cite this page

wk-concise skill. Agent Mart. https://agentmart.dev/skills/whizzzkid-wk-concise (last updated 2026-10-24)