Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Blazingly fast text search tool - recursively searches directories for regex patterns with respect to gitignore rules.
.claude/skills/penpot-ripgrep/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 19% | 0% |
Fast, smart recursive search. Respects .gitignore by default.
default code search, respects .gitignore.
bash# Search for "TODO" in current directory rg "TODO" # Case-insensitive search rg -i "fixme" # Search specific file types rg "error" -t py # Python files only rg "function" -t js # JavaScript files
bash# Whole word match rg -w "test" # Show only filenames rg -l "pattern" # Show with context (3 lines before/after) rg -C 3 "function" # Count matches rg -c "import"
bash# Multiple file types rg "error" -t py -t js # Exclude file types rg "TODO" -T md -T txt # List available types rg --type-list
bash# Regex search rg "user_\d+" # Fixed string (no regex) rg -F "function()" # Multiline search rg -U "start.*end" # Only show matches, not lines rg -o "https?://[^\s]+"
bash# Search specific directory rg "pattern" src/ # Glob patterns rg "error" -g "*.log" rg "test" -g "!*.min.js" # Include hidden files rg "secret" --hidden # Search all files (ignore .gitignore) rg "pattern" --no-ignore
bash# Preview replacements rg "old_name" --replace "new_name" # Actually replace (requires extra tool like sd) rg "old_name" -l | xargs sed -i 's/old_name/new_name/g'
bash# Parallel search (auto by default) rg "pattern" -j 8 # Skip large files rg "pattern" --max-filesize 10M # Memory map files rg "pattern" --mmap
Find TODOs in code:
bashrg "TODO|FIXME|HACK" --type-add 'code:*.{rs,go,py,js,ts}' -t code
Search in specific branches:
bashgit show branch:file | rg "pattern"
Find files containing multiple patterns:
bashrg "pattern1" | rg "pattern2"
Search with context and color:
bashrg -C 2 --color always "error" | less -R
.gitignore, skips binary filesrg is often faster than grep -r-t for file type filtering instead of --includerg pattern -l | xargs tool~/.ripgreprc--stats to see search performanceGitHub: https://github.com/BurntSushi/ripgrep User Guide: https://github.com/BurntSushi/ripgrep/blob/master/GUIDE.md
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,702 | 5,034 | -53% | 1 | 1 | 0% | 1,513 | 1,565 | +3% | 0 | 0 | — |
case-02 | pass→pass | 16,443 | 4,292 | -74% | 1 | 1 | 0% | 1,421 | 1,429 | +1% | 0 | 0 | — |
case-03 | fail→pass | 11,288 | 4,753 | -58% | 1 | 1 | 0% | 1,817 | 1,463 | -19% | 0 | 0 | — |
case-04 | fail→pass | 8,817 | 4,280 | -51% | 1 | 1 | 0% | 1,362 | 1,411 | +4% | 0 | 0 | — |
case-05 | pass→pass | 11,816 | 3,422 | -71% | 1 | 1 | 0% | 1,791 | 1,275 | -29% | 0 | 0 | — |
case-06 | pass→pass | 9,103 | 3,419 | -62% | 1 | 1 | 0% | 1,245 | 1,271 | +2% | 0 | 0 | — |
case-07 | pass→pass | 8,461 | 3,784 | -55% | 1 | 1 | 0% | 1,187 | 1,378 | +16% | 0 | 0 | — |
case-08 | pass→pass | 8,979 | 10,735 | +20% | 1 | 1 | 0% | 1,671 | 1,665 | -0% | 0 | 0 | — |
case-09 | pass→pass | 10,407 | 2,630 | -75% | 1 | 1 | 0% | 1,549 | 1,187 | -23% | 0 | 0 | — |
case-10 | fail→pass | 24,589 | 5,938 | -76% | 1 | 1 | 0% | 1,598 | 1,681 | +5% | 0 | 0 | — |
case-11 | pass→pass | 12,312 | 8,637 | -30% | 1 | 1 | 0% | 1,925 | 1,960 | +2% | 0 | 0 | — |
case-12 | pass→pass | 14,586 | 7,170 | -51% | 1 | 1 | 0% | 2,128 | 1,882 | -12% | 0 | 0 | — |
case-13 | pass→pass | 9,590 | 3,185 | -67% | 1 | 1 | 0% | 1,506 | 1,251 | -17% | 0 | 0 | — |
case-14 | pass→pass | 12,954 | 7,819 | -40% | 1 | 1 | 0% | 2,281 | 1,803 | -21% | 0 | 0 | — |
case-15 | pass→pass | 14,313 | 8,666 | -39% | 1 | 1 | 0% | 2,358 | 1,990 | -16% | 0 | 0 | — |
case-16 | fail→pass | 8,626 | 11,765 | +36% | 1 | 1 | 0% | 1,344 | 1,606 | +19% | 0 | 0 | — |
case-17 | pass→pass | 9,971 | 3,597 | -64% | 1 | 1 | 0% | 928 | 1,296 | +40% | 0 | 0 | — |
case-18 | pass→pass | 13,905 | 5,578 | -60% | 1 | 1 | 0% | 1,368 | 1,603 | +17% | 0 | 0 | — |
case-19 | pass→pass | 7,206 | 3,511 | -51% | 1 | 1 | 0% | 1,068 | 1,105 | +3% | 0 | 0 | — |
case-20 | pass→pass | 8,122 | 6,029 | -26% | 1 | 1 | 0% | 1,194 | 1,562 | +31% | 0 | 0 | — |
case-21 | fail→fail | 17,404 | 7,390 | -58% | 1 | 1 | 0% | 1,683 | 1,958 | +16% | 0 | 0 | — |
case-22 | fail→pass | 15,096 | 5,744 | -62% | 1 | 1 | 0% | 1,102 | 1,556 | +41% | 0 | 0 | — |
case-23 | fail→fail | 7,565 | 5,771 | -24% | 1 | 1 | 0% | 1,177 | 1,772 | +51% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.