flow-next-rp-explorer_skill

Overview

This skill provides token-efficient codebase exploration using the RepoPrompt CLI (rp-cli). It helps you locate, understand, and extract only the relevant code context by combining compact structural commands, AI-powered builders, and selective context exports. Use it when you want fast, low-cost code discovery and focused deep dives in Python or mixed repos.

How this skill works

The skill maps common user prompts like “use rp to explore…” to rp-cli flows that first summarize structure, then search and select relevant files, and finally export minimal context for further analysis. It prefers lightweight commands such as structure and builder to reduce token usage, isolates work into compose tabs for safe iteration, and supports chaining commands to automate select -> read -> context steps. The result is a concise context payload you can hand to other agents or use for reviews.

When to use it

• You want a quick structural overview without reading full files
• You need to find authentication, API, or data-flow code patterns
• You want to build a focused context bundle for another AI model
• You need to search and select relevant files before a deep code review
• You want token-efficient exploration to save costs or speed up feedback

Best practices

• Start with rp-cli -e 'tree' or 'structure .' to get a repo map before searching
• Use rp-cli -e 'search "term" --context-lines 2' to find hits with minimal noise
• Use builder "intent" to let the tool discover related files and open an isolated tab
• Select only matching files (select add ...) then run context to export a compact context file
• Prefer structure over reading full files and chain commands to avoid repeated large reads

Example use cases

• Investigate how authentication is implemented: structure -> search "auth" -> builder "understand authentication" -> select -> context
• Map data flow for an endpoint: structure -> search routes/controllers -> select scope -> context export for model review
• Find duplicate or similar patterns across the codebase with builder and targeted searches
• Prepare a minimal context bundle for overnight autonomous coding using Ralph mode
• Run quick pre-review discovery before a multi-model code review gate