Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use the local Gemini CLI from Claude Code with the user's existing Gemini authentication or API configuration. Use for large-context repo scans, multimodal analysis, second-opinion planning, or structured Gemini runs in the current workspace.
.claude/skills/sonichi-claude-gemini/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 133% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | 40% | 0% |
Delegate work from Claude Code to the local gemini CLI. This skill uses whatever authentication the Gemini CLI is already configured to use on this machine, including API key or signed-in CLI flows. It does not copy or export secrets.
Usage: /claude-gemini [prompt]
ARGUMENTS: $ARGUMENTS
--approval-mode plan for read-only analysis.--approval-mode auto_edit only when the user wants Gemini to make edits.--output-format json or stream-json when another tool will consume the output.bashbash "$SKILL_DIR/scripts/gemini-run.sh" --check
bash# Read-only analysis bash "$SKILL_DIR/scripts/gemini-run.sh" -- "Trace how tasks flow from voice input to execution" # Explicit model selection bash "$SKILL_DIR/scripts/gemini-run.sh" --model gemini-2.5-pro -- "Review the repo structure and identify weak points" # Machine-readable output bash "$SKILL_DIR/scripts/gemini-run.sh" --output-format json -- "Summarize risks in src/startup.sh" # Allow edits when the user asked for implementation help bash "$SKILL_DIR/scripts/gemini-run.sh" --approval-mode auto_edit -- "Implement a safer startup preflight for missing services"
--approval-mode plan for analysis.bashbash "$SKILL_DIR/scripts/gemini-run.sh" -- "$ARGUMENTS"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,572 | 19,497 | +44% | 1 | 1 | 0% | 2,110 | 754 | -64% | 0 | 0 | — |
case-02 | fail→fail | 7,773 | 6,190 | -20% | 1 | 1 | 0% | 714 | 1,254 | +76% | 0 | 0 | — |
case-03 | fail→fail | 10,736 | 6,660 | -38% | 1 | 1 | 0% | 2,145 | 853 | -60% | 0 | 0 | — |
case-04 | fail→pass | 2,959 | 3,854 | +30% | 1 | 1 | 0% | 442 | 1,030 | +133% | 0 | 0 | — |
case-05 | fail→fail | 10,870 | 5,193 | -52% | 1 | 1 | 0% | 1,875 | 781 | -58% | 0 | 0 | — |
case-06 | fail→fail | 6,350 | 5,305 | -16% | 1 | 1 | 0% | 983 | 678 | -31% | 0 | 0 | — |
case-07 | fail→fail | 12,622 | 7,423 | -41% | 1 | 1 | 0% | 1,959 | 802 | -59% | 0 | 0 | — |
case-08 | fail→fail | 11,861 | 6,985 | -41% | 1 | 1 | 0% | 2,028 | 951 | -53% | 0 | 0 | — |
case-09 | fail→fail | 2,742 | 5,569 | +103% | 1 | 1 | 0% | 297 | 762 | +157% | 0 | 0 | — |
case-10 | fail→fail | 12,907 | 6,742 | -48% | 1 | 1 | 0% | 2,358 | 900 | -62% | 0 | 0 | — |
case-11 | fail→pass | 8,166 | 2,618 | -68% | 1 | 1 | 0% | 1,195 | 880 | -26% | 0 | 0 | — |
case-12 | fail→pass | 3,361 | 7,140 | +112% | 1 | 1 | 0% | 427 | 843 | +97% | 0 | 0 | — |
case-13 | fail→fail | 10,883 | 5,627 | -48% | 1 | 1 | 0% | 1,907 | 768 | -60% | 0 | 0 | — |
case-14 | fail→pass | 10,810 | 1,569 | -85% | 1 | 1 | 0% | 1,983 | 692 | -65% | 0 | 0 | — |
case-15 | fail→fail | 17,533 | 6,447 | -63% | 1 | 1 | 0% | 2,863 | 745 | -74% | 0 | 0 | — |
case-16 | fail→fail | 2,501 | 6,347 | +154% | 1 | 1 | 0% | 302 | 782 | +159% | 0 | 0 | — |
case-17 | fail→fail | 4,981 | 6,453 | +30% | 1 | 1 | 0% | 730 | 687 | -6% | 0 | 0 | — |
case-18 | fail→fail | 17,967 | 24,207 | +35% | 1 | 1 | 0% | 1,615 | 752 | -53% | 0 | 0 | — |
case-19 | fail→fail | 13,786 | 6,532 | -53% | 1 | 1 | 0% | 2,313 | 808 | -65% | 0 | 0 | — |
case-20 | pass→pass | 8,584 | 8,411 | -2% | 1 | 1 | 0% | 1,555 | 2,177 | +40% | 0 | 0 | — |
case-21 | pass→pass | 2,432 | 1,906 | -22% | 1 | 1 | 0% | 353 | 668 | +89% | 0 | 0 | — |
case-22 | pass→pass | 4,819 | 8,506 | +77% | 1 | 1 | 0% | 752 | 1,431 | +90% | 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 8 counted toward the lift figure. The other 14 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 +18 percentage points is the difference between those two pass rates over the 8 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.