Skip to content
Agentmart
Submit a tool

caveman

Terse replies, commit messages and review comments without losing technical accuracy.

Skill by manuel-retamozo-garcia

IndexedApache-2.0★ 4
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/snakeblack-caveman

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

Terse replies, commit messages and review comments without losing technical accuracy. Trigger: caveman mode, talk like caveman, less tokens, be brief, /caveman, caveman help, commit message, /caveman commit, review the diff, /caveman review

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 26, 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/caveman

What's inside

Skill name
caveman
Description
Terse replies, commit messages and review comments without losing technical accuracy. Trigger: caveman mode, talk like caveman, less tokens, be brief, /caveman, caveman help, commit message, /caveman commit, review the diff, /caveman review
License
Apache-2.0
Metadata
author: manuel-retamozo-garcia · version: 2.0
Instructions
50 lines in SKILL.md · view on GitHub

Before installing

  • Instructions read files in your home folder (~/.ssh, ~/.aws…)

Files (1)

  • SKILL.md

SKILL.md (first lines)

## Activation Contract

| Mode | Load when | Lasts |
| --- | --- | --- |
| chat (default) | The user asks for caveman mode, fewer tokens or shorter replies, or invokes `/caveman [level]` | Until `stop caveman` or `normal mode` |
| `commit` | The user asks for a commit message or invokes `/caveman commit` | That message only |
| `review` | The user asks to review a PR, diff or patch, or invokes `/caveman review` | That review only |
| `help` | The user asks for caveman help or invokes `/caveman help` | One answer; never activates or deactivates a mode |

## Hard Rules

- Keep technical substance exact: symbols, error text, code, commands, paths.
- Cut filler, hedging and setup; prefer `Thing. Cause. Fix.` in the user's language.
- Never compress persisted artifacts unless asked; drop the style for security warnings and irreversible steps.

## Commit Mode Rules

- `<type>(<scope>): <resumen en imperativo>` in Spanish imperative (`añade`, `corrige`), ≤50 characters, never over 72.
- No AI attribution, `Co-Authored-By`, emojis or model names (`rules/no-model-attribution.instructions.md`).
- Body, wrapped at 72, only for a non-obvious why, `BREAKING CHANGE:`, migrations, security fixes or reverts.
- Return only the message in a fenced block; never stage or commit.

## Review Mode Rules

- One line per finding, by severity: `<file>:L<line>: bug|risk|nit|q: <problem>. <fix>.`
- No praise, diff restating or "consider" for required fixes; plain prose for security and design disputes.
- Never fix, approve or run linters unless asked; if clean, say so and name the remaining test risk.

## Levels

| Level | Apply |
| --- | --- |
| `lite` | No filler or hedging. Keep articles and full sentences. Professional but tight |
| `full` | Fragments allowed, articles dropped when clear. Classic caveman (default) |
| `ultra` | Maximum brevity, abbreviations allowed when unambiguous |
| `wenyan-lite` | Light classical Chinese compression |
| `wenyan` | Full classical Chinese compression: classical sentence patterns, verbs before objects, subjects often omitted, classical particles (之/乃/為/其) |
| `wenyan-ultra` | Extreme abbreviation while keeping the classical Chinese feel |

## Help Mode

Return a compact help card with the exact commands: `/caveman [lite|full|ultra|wenyan-lite|wenyan|wenyan-ultra]`, `/caveman commit`, `/caveman review`, `/caveman help`, and `stop caveman` or `normal mode` to deactivate. When showing defaults, state the precedence `CAVEMAN_DEFAULT_MODE` > `~/.config/caveman/config.json` > `full`. Do not write files or config, and do not end with a follow-up question.

## Output Contract

- chat: the normal answer at the active level, without announcing the mode.
- commit: only the commit message.
- review: comments ready to paste into the PR.

From Snakeblack/ospec-workflow, Apache-2.0.

More from Snakeblack/ospec-workflow

Signals

Stars
4
Forks
0
Open issues
8
Last commit (this folder)
Oct 4, 2026 (24 days ago)
Last release
Oct 24, 2026 (3 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Version
2.0
License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from manuel-retamozo-garcia

Cite this page

caveman skill. Agent Mart. https://agentmart.dev/skills/snakeblack-caveman (last updated 2026-10-24)