Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Semantic code search using mgrep for efficient codebase exploration. This skill should be used when searching or exploring codebases with more than 30 non-gitignored files and/or nested directory structures. It provides natural language semantic search that complements traditional grep/ripgrep for finding features, understanding intent, and exploring unfamiliar code.
.claude/skills/mkurman-mgrep-code-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -44% | 0% |
--|-------------| | -m <count> | Maximum results (default: 10) | | -c, --content | Display full result content | | -a, --answer | Generate AI-powered synthesis of results | | -s, --sync | Update index before searching | | --no-rerank | Disable relevance optimisation |
bash# Get more results bunx @mixedbread/mgrep -m 25 "user authentication flow" # Show full content of matches bunx @mixedbread/mgrep -c "error handling patterns" # Get an AI-synthesised answer bunx @mixedbread/mgrep -a "how does the caching layer work?" # Sync index before searching bunx @mixedbread/mgrep -s "payment processing" src/services
bash bunx @mixedbread/mgrep watch
bash bunx @mixedbread/mgrep "what you're looking for" [optional/path]
bash bunx @mixedbread/mgrep -m 20 -c "refined query" src/specific/directory
Configure defaults via environment variables:
| Variable | Purpose | |----------|---------| | MGREP_MAX_COUNT | Default result limit | | MGREP_CONTENT | Enable content display (1/true) | | MGREP_ANSWER | Enable AI synthesis (1/true) | | MGREP_SYNC | Pre-search sync (1/true) |
bunx @mixedbread/mgrep to run commands (not npm/npx or direct installation)bunx @mixedbread/mgrep watch before searching to ensure the index is current.gitignore patterns automatically.mgrepignore for additional exclusions| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-22 | pass→pass | 10,919 | 7,523 | -31% | 1 | 1 | 0% | 1,939 | 1,786 | -8% | 0 | 0 | — |
case-01 | fail→pass | 9,192 | 1,328 | -86% | 1 | 1 | 0% | 1,559 | 686 | -56% | 0 | 0 | — |
case-02 | fail→pass | 25,354 | 3,688 | -85% | 1 | 1 | 0% | 1,924 | 1,053 | -45% | 0 | 0 | — |
case-03 | fail→pass | 13,184 | 8,192 | -38% | 1 | 1 | 0% | 1,527 | 1,836 | +20% | 0 | 0 | — |
case-04 | fail→pass | 11,015 | 2,648 | -76% | 1 | 1 | 0% | 1,723 | 813 | -53% | 0 | 0 | — |
case-05 | pass→pass | 12,314 | 3,517 | -71% | 1 | 1 | 0% | 1,649 | 789 | -52% | 0 | 0 | — |
case-06 | fail→pass | 8,323 | 2,387 | -71% | 1 | 1 | 0% | 1,401 | 782 | -44% | 0 | 0 | — |
case-07 | fail→pass | 16,262 | 2,362 | -85% | 1 | 1 | 0% | 2,817 | 911 | -68% | 0 | 0 | — |
case-08 | fail→pass | 34,036 | 2,208 | -94% | 1 | 1 | 0% | 6,025 | 819 | -86% | 0 | 0 | — |
case-09 | fail→pass | 14,302 | 3,458 | -76% | 1 | 1 | 0% | 2,161 | 996 | -54% | 0 | 0 | — |
case-10 | fail→pass | 15,445 | 2,191 | -86% | 1 | 1 | 0% | 2,607 | 767 | -71% | 0 | 0 | — |
case-11 | fail→pass | 9,248 | 1,711 | -81% | 1 | 1 | 0% | 1,497 | 756 | -49% | 0 | 0 | — |
case-12 | fail→pass | 7,238 | 5,171 | -29% | 1 | 1 | 0% | 362 | 1,252 | +246% | 0 | 0 | — |
case-13 | fail→pass | 5,806 | 2,245 | -61% | 1 | 1 | 0% | 901 | 811 | -10% | 0 | 0 | — |
case-14 | fail→fail | 29,801 | 4,945 | -83% | 1 | 1 | 0% | 6,171 | 674 | -89% | 0 | 0 | — |
case-15 | pass→pass | 10,523 | 2,634 | -75% | 1 | 1 | 0% | 1,686 | 925 | -45% | 0 | 0 | — |
case-16 | fail→pass | 9,193 | 1,902 | -79% | 1 | 1 | 0% | 1,376 | 735 | -47% | 0 | 0 | — |
case-17 | fail→pass | 12,914 | 1,948 | -85% | 1 | 1 | 0% | 2,123 | 771 | -64% | 0 | 0 | — |
case-18 | fail→pass | 9,386 | 1,688 | -82% | 1 | 1 | 0% | 1,418 | 734 | -48% | 0 | 0 | — |
case-19 | fail→pass | 8,877 | 2,472 | -72% | 1 | 1 | 0% | 1,286 | 897 | -30% | 0 | 0 | — |
case-20 | pass→pass | 8,044 | 5,539 | -31% | 1 | 1 | 0% | 1,329 | 1,342 | +1% | 0 | 0 | — |
case-21 | pass→pass | 8,411 | 4,924 | -41% | 1 | 1 | 0% | 1,428 | 1,241 | -13% | 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. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +73 percentage points is the difference between those two pass rates over the 20 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.