rg

name: rg description: "⚡ IMPORTANT: Auto-trigger for text search. Use when: search for text, find files containing, locate TODOs, security scan secrets, trace class usage, grep pattern matching. Returns only matching lines. 30-98% token savings. 3-4x faster than grep/Select-String."

Text search with ripgrep (rg)

Core rule

When you need to search for text content inside files, use rg on the command line instead of Select-String (PowerShell), grep -r (Bash), or reading entire files.

rg (ripgrep) is a fast regex-aware search tool that respects .gitignore, outputs relative paths, and produces compact results. For an AI agent, this matters because:

• Extracts only matching lines: Instead of loading a 2.6 MB JSON file (692k tokens) and searching in-context, rg returns just the 415 matching lines (6k tokens) — a 98% reduction.
• Relative paths reduce overhead: Same benefit as fd — shorter paths = fewer tokens per match.
• .gitignore awareness: Automatically skips build artifacts, dependencies, and generated code that would add noise and consume tokens without providing value.
• Speed at scale: 3-4x faster than Select-String on 10k+ files means results arrive sooner, and the model can act on them faster.

Prerequisites: check/install ripgrep

IMPORTANT: Check availability first:

rg --version

If rg is not installed:

Shell compatibility

rg works natively in both Bash and PowerShell — no wrapping needed:

rg "groot/sa-id" catalog/systems

rg "groot/sa-id" catalog\systems

Usage patterns

Search for a string

rg "agreement_id" event.json
rg "spec.system" catalog/components

rg "agreement_id" event.json
rg "spec.system" catalog\components

List files containing a pattern (no content)

rg -l "type: repository" catalog/components
# → Just the file paths, one per line

rg -l "groot-sa-majeur" catalog/systems | wc -l
# → 749

(rg -l "groot-sa-majeur" catalog\systems).Count

Count matches per file

rg -c "agreement_id" event.json
# → event.json:415

Search with regex

rg "sa-\d{4}-" catalog/systems --no-filename -o | sort -u
rg "owner: .*team" catalog/components -l

Extract specific lines (like grep -o)

rg "name: (.*)" catalog/systems --no-filename -o
# Extracts just the matching lines, no file paths

rg "name: (.*)" catalog\systems --no-filename -o

Search with context lines

rg "spec.system" catalog/components/af-ose -A 2 -B 2
# Shows 2 lines before and after each match

Restrict to specific file types

rg "TODO" -t yaml catalog
rg "agreement_id" -t json .project

Include gitignored files

rg "pattern" --no-ignore              # Skip .gitignore rules
rg "pattern" -uuu                     # Unrestricted (all files)

Replace text (preview)

rg "old-system" catalog/components -r "new-system"
# Shows what replacements would look like (does NOT modify files)

Combine results with other commands

# Count components per project directory
rg -l "type: repository" catalog/components | sed 's|/[^/]*$||' | sort | uniq -c | sort -rn

# PowerShell equivalent
rg -l "type: repository" catalog\components | Group-Object { Split-Path (Split-Path $_) -Leaf } | Sort-Object Count -Descending