Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Work with files and directories mounted in a Mirage virtual filesystem. Use when a task mentions Mirage, mounted cloud or database data, Mirage virtual paths, or asks to inspect, search, create, or edit data exposed through the Mirage tools.
.claude/skills/strukto-ai-mirage-filesystem/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -68% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -69% | 0% |
Use the Mirage tools for virtual paths. Host filesystem tools cannot access those paths unless the user separately configured a FUSE mount.
ls, grep, or execute_command tool to discover mounted data.read tool before modifying an existing file.edit for an existing file and write only for a new file.execute_command for pipelines and structured-file commands that need Mirage shell semantics.Do not fall back to a host filesystem tool when a Mirage tool fails on a virtual path. Report the Mirage error or fix the Mirage configuration.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,449 | 4,558 | -29% | 1 | 1 | 0% | 1,337 | 527 | -61% | 0 | 0 | — |
case-02 | fail→fail | 15,327 | 3,252 | -79% | 1 | 1 | 0% | 2,766 | 362 | -87% | 0 | 0 | — |
case-03 | fail→fail | 8,599 | 5,583 | -35% | 1 | 1 | 0% | 1,756 | 429 | -76% | 0 | 0 | — |
case-04 | pass→fail | 6,597 | 3,889 | -41% | 1 | 1 | 0% | 1,378 | 424 | -69% | 0 | 0 | — |
case-05 | pass→pass | 2,619 | 2,328 | -11% | 1 | 1 | 0% | 459 | 610 | +33% | 0 | 0 | — |
case-06 | pass→fail | 5,709 | 4,172 | -27% | 1 | 1 | 0% | 1,083 | 401 | -63% | 0 | 0 | — |
case-07 | pass→pass | 3,191 | 2,192 | -31% | 1 | 1 | 0% | 534 | 534 | 0% | 0 | 0 | — |
case-08 | pass→pass | 5,654 | 2,758 | -51% | 1 | 1 | 0% | 953 | 574 | -40% | 0 | 0 | — |
case-09 | fail→pass | 6,847 | 2,296 | -66% | 1 | 1 | 0% | 1,224 | 549 | -55% | 0 | 0 | — |
case-10 | pass→pass | 7,998 | 3,130 | -61% | 1 | 1 | 0% | 1,327 | 717 | -46% | 0 | 0 | — |
case-11 | fail→pass | 12,814 | 3,637 | -72% | 1 | 1 | 0% | 2,332 | 838 | -64% | 0 | 0 | — |
case-12 | pass→fail | 10,029 | 5,081 | -49% | 1 | 1 | 0% | 1,975 | 492 | -75% | 0 | 0 | — |
case-17 | fail→pass | 7,208 | 2,569 | -64% | 1 | 1 | 0% | 1,113 | 650 | -42% | 0 | 0 | — |
case-13 | pass→pass | 6,027 | 2,315 | -62% | 1 | 1 | 0% | 1,122 | 546 | -51% | 0 | 0 | — |
case-14 | pass→pass | 9,435 | 2,084 | -78% | 1 | 1 | 0% | 1,697 | 571 | -66% | 0 | 0 | — |
case-15 | pass→pass | 8,969 | 5,780 | -36% | 1 | 1 | 0% | 1,634 | 667 | -59% | 0 | 0 | — |
case-16 | fail→pass | 7,448 | 1,590 | -79% | 1 | 1 | 0% | 1,355 | 427 | -68% | 0 | 0 | — |
case-18 | pass→pass | 9,559 | 4,180 | -56% | 1 | 1 | 0% | 1,474 | 809 | -45% | 0 | 0 | — |
case-19 | pass→pass | 14,130 | 3,426 | -76% | 1 | 1 | 0% | 2,220 | 631 | -72% | 0 | 0 | — |
case-20 | pass→fail | 9,144 | 7,815 | -15% | 1 | 1 | 0% | 1,687 | 732 | -57% | 0 | 0 | — |
case-21 | pass→pass | 2,215 | 1,490 | -33% | 1 | 1 | 0% | 372 | 377 | +1% | 0 | 0 | — |
case-22 | pass→pass | 12,685 | 2,641 | -79% | 1 | 1 | 0% | 2,009 | 605 | -70% | 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 15 counted toward the lift figure. The other 7 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 0 percentage points is the difference between those two pass rates over the 15 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
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.