Skip to content
Agentmart
Submit a tool

token-efficient

Use when the user activates compressed output mode ("caveman mode", "terse", "compress") — drops conversational filler while preserving artifact quality.

Skill by ovargas

IndexedMIT★ 14
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/ovargas-token-efficient

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

Use when the user activates compressed output mode ("caveman mode", "terse", "compress") — drops conversational filler while preserving artifact quality. Deactivates on "normal mode" or "verbose". ADR-003 scopes this to conversation only.

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 and code. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/token-efficient

What's inside

Skill name
token-efficient
Description
Use when the user activates compressed output mode ("caveman mode", "terse", "compress") — drops conversational filler while preserving artifact quality. Deactivates on "normal mode" or "verbose". ADR-003 scopes this to conversation only.
License
MIT
Instructions
65 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Token-Efficient Communication Mode

## Activation

**On:** User says "caveman mode", "terse mode", "terse", "compress", "be brief", or similar.
**Off:** User says "normal mode", "verbose", "full prose", or similar.

Mode persists for the rest of the session or until deactivated. Acknowledge activation/deactivation in one short line:

- On: `Caveman mode on.`
- Off: `Normal mode restored.`

## Compression Rules

When active, apply these rules to all **conversational output** (status updates, explanations, confirmations, progress reports, inline commentary):

1. **Drop articles** (a, an, the) where meaning is preserved
2. **Drop filler phrases** ("it is important to note", "in order to", "let me")
3. **Drop pleasantries** ("Great question!", "Sure thing!", "Happy to help")
4. **Drop hedging** ("I think", "it seems like", "probably", "might")
5. **Fragments OK** — no need for complete sentences in status updates
6. **Technical terms exact** — never abbreviate domain vocabulary, paths, identifiers, or numbers
7. **Tool calls unchanged** — arguments to Edit, Write, Bash, etc. are never compressed

## What Is NEVER Compressed (ADR-003)

This list is not configurable. These outputs use full prose regardless of mode:

- Feature specs, implementation plans, reviews, bug reports
- PR descriptions and commit messages
- ADRs, CONTEXT.md, backlog entries
- Any content written via Write or Edit tools (file content is an artifact)
- Any output that persists beyond the current session

**Rule of thumb:** If it goes into a file or a git object, it is an artifact. Full prose.

## Auto-Clarity Exception

Compression automatically suspends (full prose resumes) for:

- **Security warnings** — vulnerabilities, credential exposure, permission issues
- **Irreversible action confirmations** — destructive git operations, file deletions, database drops
- **Error explanations** — when diagnosing a failure, precision matters more than brevity
- **Multi-step instructions** — when compressed language risks the user misunderstanding the sequence

After the exception context passes, compression resumes without user action. Do not announce the switch.

## Examples

**Without mode (normal):**
> I've read the feature spec and identified the files that need to change. The main modification will be to the user service, where we need to add the validation logic. I'll also need to update the corresponding test file. Let me start with the test first, following TDD.

**With mode (caveman):**
> Read spec. Changes: user service (add validation), test file. Starting with test (TDD).

**Auto-Clarity Exception (even in caveman mode):**
> WARNING: This will force-push to main, overwriting 3 commits from other contributors. This cannot be undone. Proceed?

## Integration

- **Triggered by:** User activation phrase (mapped in `skills/skill-awareness/SKILL.md`)
- **Complements:** `skills/humanizer/SKILL.md` (humanizer polishes artifacts; this compresses conversation)
- **Constraint:** ADR-003 (`docs/decisions/2026-04-29-token-mode-scope.md`)

From ovargas/virtual-team, MIT.

More from ovargas/virtual-team

Signals

Stars
14
Forks
1
Open issues
0
Last commit (this folder)
May 12, 2026 (5 months ago)
Last release
Oct 8, 2026 (20 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from ovargas

Cite this page

token-efficient skill. Agent Mart. https://agentmart.dev/skills/ovargas-token-efficient (last updated 2026-10-24)