cavecrew

Determine when to use subagents for code investigations, edits, and reviews.

Updated Jun 21, 2026
One-click install
npx skills add https://github.com/Bonny94ITA/BottegaWeb --skill cavecrew-bonny94ita
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: cavecrew
Source: https://github.com/Bonny94ITA/BottegaWeb/tree/main/.claude/skills/cavecrew
Command: npx skills add https://github.com/Bonny94ITA/BottegaWeb --skill cavecrew-bonny94ita

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Helps manage context exhaustion during long sessions by using compressed subagent output to reduce the amount of context needed for tool results.

Core Features & Use Cases

  • Subagent Decision Guide: Determines when to use subagents like cavecrew-investigator, cavecrew-builder, and cavecrew-reviewer.
  • Compressed Output: Subagent results use 1/3 the tokens of vanilla prose, conserving context for longer sessions.
  • Use Case: When working with complex codebases and need to perform locate, edit, and review tasks, this skill helps maintain session context.

Quick Start

Use the cavecrew skill to locate code references for a function by typing 'locate function_name'.

Frequently Asked Questions about cavecrew

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent context exhaustion during long code editing sessions in large codebases?

To prevent context exhaustion during long code editing sessions, you can use compressed subagent outputs to reduce the amount of context needed for tool results, conserving tokens for longer work.

When should I use subagents for code investigation and review tasks?

You should use subagents for code investigation and review tasks when working with complex codebases and you need to perform locate, edit, and review operations while maintaining session context.

How do I locate function references in a complex codebase without running out of context?

You can locate function references without running out of context by using subagents like cavecrew-investigator, which compresses the tool results to use one-third of the tokens compared to vanilla prose.

What is the best way to optimize context usage when performing code reviews?

The best way to optimize context usage when performing code reviews is to delegate the task to a reviewer subagent that compresses its output, using only one-third of the tokens required by vanilla prose.

Does compressed subagent output work for both code editing and code investigation?

Yes, compressed subagent output works for both code editing and investigation, utilizing specialized subagents like cavecrew-builder and cavecrew-investigator to minimize context token usage across tasks.