Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Deploy, configure, and integrate Sandbox Agent - a universal API for orchestrating AI coding agents (Claude Code, Codex, OpenCode, Amp) in sandboxed environments. Use when setting up sandbox-agent server locally or in cloud sandboxes (E2B, Daytona, Docker), creating and managing agent sessions via SDK or API, streaming agent events and handling human-in-the-loop interactions, building chat UIs for coding agents, or understanding the universal schema for agent responses.
.claude/skills/rivet-dev-sandbox-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -61% | 0% |
Sandbox Agent provides a universal API for orchestrating AI coding agents in sandboxed environments.
If something is not working as intended or you are stuck, prompt the user to join the Rivet Discord or file an issue on GitHub to report an issue and get help.
Provide the user with a pre-generated report with:
{{QUICKSTART}}
{{REFERENCE_MAP}}
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | fail→pass | 10,792 | 3,923 | -64% | 1 | 1 | 0% | 1,921 | 1,007 | -48% | 0 | 0 | — |
case-18 | fail→pass | 8,993 | 4,085 | -55% | 1 | 1 | 0% | 1,607 | 956 | -41% | 0 | 0 | — |
case-01 | fail→pass | 10,440 | 4,861 | -53% | 1 | 1 | 0% | 1,941 | 1,151 | -41% | 0 | 0 | — |
case-02 | fail→pass | 15,934 | 5,520 | -65% | 1 | 1 | 0% | 2,339 | 1,250 | -47% | 0 | 0 | — |
case-03 | fail→fail | 10,369 | 5,468 | -47% | 1 | 1 | 0% | 1,858 | 1,315 | -29% | 0 | 0 | — |
case-04 | fail→pass | 6,018 | 2,052 | -66% | 1 | 1 | 0% | 1,299 | 511 | -61% | 0 | 0 | — |
case-05 | fail→pass | 5,635 | 2,354 | -58% | 1 | 1 | 0% | 969 | 572 | -41% | 0 | 0 | — |
case-06 | fail→pass | 8,087 | 3,250 | -60% | 1 | 1 | 0% | 1,337 | 505 | -62% | 0 | 0 | — |
case-07 | pass→pass | 7,385 | 2,557 | -65% | 1 | 1 | 0% | 1,216 | 648 | -47% | 0 | 0 | — |
case-08 | pass→pass | 8,122 | 3,049 | -62% | 1 | 1 | 0% | 1,426 | 537 | -62% | 0 | 0 | — |
case-09 | pass→pass | 7,301 | 2,534 | -65% | 1 | 1 | 0% | 1,256 | 671 | -47% | 0 | 0 | — |
case-10 | pass→pass | 7,825 | 1,571 | -80% | 1 | 1 | 0% | 1,441 | 463 | -68% | 0 | 0 | — |
case-11 | pass→pass | 9,348 | 2,036 | -78% | 1 | 1 | 0% | 1,720 | 544 | -68% | 0 | 0 | — |
case-12 | fail→pass | 11,080 | 1,475 | -87% | 1 | 1 | 0% | 2,172 | 490 | -77% | 0 | 0 | — |
case-13 | pass→pass | 14,051 | 1,502 | -89% | 1 | 1 | 0% | 2,226 | 460 | -79% | 0 | 0 | — |
case-14 | fail→pass | 8,275 | 5,739 | -31% | 1 | 1 | 0% | 1,525 | 1,302 | -15% | 0 | 0 | — |
case-15 | fail→pass | 10,892 | 4,902 | -55% | 1 | 1 | 0% | 2,109 | 1,181 | -44% | 0 | 0 | — |
case-16 | fail→fail | 12,535 | 4,276 | -66% | 1 | 1 | 0% | 2,245 | 1,090 | -51% | 0 | 0 | — |
case-17 | fail→pass | 13,888 | 4,764 | -66% | 1 | 1 | 0% | 2,412 | 1,023 | -58% | 0 | 0 | — |
case-20 | pass→pass | 5,005 | 2,874 | -43% | 1 | 1 | 0% | 924 | 745 | -19% | 0 | 0 | — |
case-21 | pass→pass | 12,997 | 6,177 | -52% | 1 | 1 | 0% | 2,721 | 1,491 | -45% | 0 | 0 | — |
case-22 | pass→pass | 6,774 | 5,612 | -17% | 1 | 1 | 0% | 1,283 | 1,266 | -1% | 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. The headline lift of +50 percentage points is the difference between those two pass rates over the 22 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.