Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Persist and restore agent sessions across conversations with state snapshots
.claude/skills/ruvnet-session-persist/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -43% | 0% |
Save and restore complete agent sessions across conversations.
When you need to pause work and resume later with full context, or when you want to checkpoint progress during long-running tasks.
mcp__plugin_ruflo-core_ruflo__session_save to snapshot current statemcp__plugin_ruflo-core_ruflo__session_list to see all saved sessionsmcp__plugin_ruflo-core_ruflo__session_restore to resume a previous sessionmcp__plugin_ruflo-core_ruflo__session_info for session details and metadatamcp__plugin_ruflo-core_ruflo__session_delete to remove old sessionshooks_session-start — automatically restore context at conversation starthooks_session-end — automatically save state with metrics exporthooks_session-restore — restore a specific session by ID| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 4,406 | 2,077 | -53% | 1 | 1 | 0% | 905 | 639 | -29% | 0 | 0 | — |
case-11 | fail→fail | 7,282 | 3,303 | -55% | 1 | 1 | 0% | 1,328 | 813 | -39% | 0 | 0 | — |
case-01 | fail→fail | 9,131 | 3,082 | -66% | 1 | 1 | 0% | 1,519 | 747 | -51% | 0 | 0 | — |
case-02 | fail→fail | 3,429 | 1,607 | -53% | 1 | 1 | 0% | 503 | 476 | -5% | 0 | 0 | — |
case-03 | fail→fail | 1,732 | 1,569 | -9% | 1 | 1 | 0% | 265 | 560 | +111% | 0 | 0 | — |
case-04 | pass→pass | 5,472 | 5,760 | +5% | 1 | 1 | 0% | 1,068 | 1,262 | +18% | 0 | 0 | — |
case-06 | pass→pass | 3,637 | 3,294 | -9% | 1 | 1 | 0% | 674 | 854 | +27% | 0 | 0 | — |
case-07 | fail→pass | 9,752 | 1,743 | -82% | 1 | 1 | 0% | 1,618 | 621 | -62% | 0 | 0 | — |
case-08 | pass→pass | 9,674 | 1,682 | -83% | 1 | 1 | 0% | 1,667 | 486 | -71% | 0 | 0 | — |
case-09 | fail→pass | 15,513 | 2,951 | -81% | 1 | 1 | 0% | 2,624 | 819 | -69% | 0 | 0 | — |
case-10 | fail→fail | 5,806 | 2,252 | -61% | 1 | 1 | 0% | 966 | 702 | -27% | 0 | 0 | — |
case-12 | fail→fail | 4,876 | 1,629 | -67% | 1 | 1 | 0% | 851 | 565 | -34% | 0 | 0 | — |
case-13 | pass→pass | 11,271 | 3,014 | -73% | 1 | 1 | 0% | 2,010 | 704 | -65% | 0 | 0 | — |
case-14 | pass→pass | 12,260 | 8,045 | -34% | 1 | 1 | 0% | 1,906 | 635 | -67% | 0 | 0 | — |
case-15 | pass→pass | 11,177 | 2,502 | -78% | 1 | 1 | 0% | 1,824 | 671 | -63% | 0 | 0 | — |
case-21 | fail→pass | 7,301 | 2,029 | -72% | 1 | 1 | 0% | 1,184 | 651 | -45% | 0 | 0 | — |
case-16 | pass→pass | 12,157 | 2,675 | -78% | 1 | 1 | 0% | 2,048 | 762 | -63% | 0 | 0 | — |
case-17 | fail→pass | 4,830 | 1,968 | -59% | 1 | 1 | 0% | 878 | 639 | -27% | 0 | 0 | — |
case-18 | fail→pass | 7,045 | 2,125 | -70% | 1 | 1 | 0% | 1,223 | 696 | -43% | 0 | 0 | — |
case-19 | fail→pass | 6,340 | 1,667 | -74% | 1 | 1 | 0% | 1,200 | 563 | -53% | 0 | 0 | — |
case-20 | fail→pass | 6,933 | 2,400 | -65% | 1 | 1 | 0% | 1,062 | 788 | -26% | 0 | 0 | — |
case-22 | fail→pass | 7,751 | 2,220 | -71% | 1 | 1 | 0% | 1,345 | 675 | -50% | 0 | 0 | — |
case-23 | fail→pass | 9,330 | 1,715 | -82% | 1 | 1 | 0% | 1,540 | 518 | -66% | 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 +39 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.