Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Recover session state from memory-keeper after context loss.
.claude/skills/aiskillstore-recovery/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -51% | 0% |
Restore session state from memory-keeper checkpoints after context is lost or when starting a new session.
/clear or session restartcontext_get(limit: 50, sort: "created_desc")context_summarize()context_get(category: "progress", limit: 20)context_get(priority: "high", limit: 10)From the retrieved context, extract:
current-task: What was being worked onfiles-modified: Files that were changedimplementation-progress: How far alongnext-action: What needs to happen nextblockers: Any known issuesPresent the recovered state clearly:
markdown## Session Recovered ### Previous Task <current-task value> ### Progress <implementation-progress value> ### Files Modified <list of files-modified> ### Blockers/Issues <any blockers found> ### Recommended Next Action <next-action value> ### Recent Checkpoints 1. <checkpoint 1 name>: <description> 2. <checkpoint 2 name>: <description> --- Ready to continue. Confirm to proceed with: <next-action>
If memory-keeper has no relevant context:
markdown## No Previous Context Found No checkpoints or progress items found in memory-keeper. Possible reasons: - This is a new session with no prior work - Previous session did not checkpoint (work may be lost) - Memory was cleared To start fresh, describe what you'd like to work on.
If only some context is found:
markdown## Partial Recovery Found limited context from previous session: ### Available Information <whatever was found> ### Missing Information - [ ] Current task (not found) - [ ] Files modified (not found) - etc. Would you like to: 1. Continue with available context 2. Start fresh 3. Provide additional context manually
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,343 | 17,820 | +44% | 1 | 1 | 0% | 486 | 693 | +43% | 0 | 0 | — |
case-02 | fail→fail | 10,377 | 10,219 | -2% | 1 | 1 | 0% | 826 | 2,399 | +190% | 0 | 0 | — |
case-03 | fail→fail | 4,087 | 14,147 | +246% | 1 | 1 | 0% | 501 | 1,016 | +103% | 0 | 0 | — |
case-04 | fail→pass | 9,274 | 7,312 | -21% | 1 | 1 | 0% | 1,388 | 1,016 | -27% | 0 | 0 | — |
case-05 | fail→pass | 16,582 | 7,427 | -55% | 1 | 1 | 0% | 1,843 | 1,038 | -44% | 0 | 0 | — |
case-06 | fail→pass | 14,646 | 4,321 | -70% | 1 | 1 | 0% | 2,533 | 1,318 | -48% | 0 | 0 | — |
case-07 | fail→pass | 15,519 | 3,580 | -77% | 1 | 1 | 0% | 1,564 | 1,153 | -26% | 0 | 0 | — |
case-08 | pass→pass | 17,318 | 10,197 | -41% | 1 | 1 | 0% | 1,746 | 1,332 | -24% | 0 | 0 | — |
case-09 | fail→fail | 18,417 | 7,596 | -59% | 1 | 1 | 0% | 2,075 | 1,083 | -48% | 0 | 0 | — |
case-10 | fail→fail | 8,496 | 3,405 | -60% | 1 | 1 | 0% | 1,518 | 959 | -37% | 0 | 0 | — |
case-11 | fail→fail | 11,705 | 3,918 | -67% | 1 | 1 | 0% | 893 | 1,267 | +42% | 0 | 0 | — |
case-12 | pass→pass | 19,047 | 15,290 | -20% | 1 | 1 | 0% | 2,347 | 2,071 | -12% | 0 | 0 | — |
case-13 | fail→pass | 11,775 | 6,832 | -42% | 1 | 1 | 0% | 2,015 | 983 | -51% | 0 | 0 | — |
case-14 | fail→pass | 17,549 | 7,874 | -55% | 1 | 1 | 0% | 2,106 | 1,049 | -50% | 0 | 0 | — |
case-15 | fail→pass | 19,485 | 4,640 | -76% | 1 | 1 | 0% | 2,208 | 1,196 | -46% | 0 | 0 | — |
case-16 | pass→fail | 11,164 | 7,482 | -33% | 1 | 1 | 0% | 940 | 1,024 | +9% | 0 | 0 | — |
case-17 | pass→pass | 9,242 | 2,376 | -74% | 1 | 1 | 0% | 1,761 | 1,036 | -41% | 0 | 0 | — |
case-18 | fail→pass | 11,839 | 6,885 | -42% | 1 | 1 | 0% | 1,852 | 872 | -53% | 0 | 0 | — |
case-19 | pass→pass | 16,655 | 2,759 | -83% | 1 | 1 | 0% | 1,595 | 873 | -45% | 0 | 0 | — |
case-20 | fail→fail | 15,652 | 2,842 | -82% | 1 | 1 | 0% | 1,436 | 868 | -40% | 0 | 0 | — |
case-21 | fail→fail | 22,622 | 9,912 | -56% | 1 | 1 | 0% | 1,273 | 978 | -23% | 0 | 0 | — |
case-22 | fail→pass | 10,079 | 9,151 | -9% | 1 | 1 | 0% | 1,621 | 1,272 | -22% | 0 | 0 | — |
case-23 | fail→fail | 14,610 | 7,414 | -49% | 1 | 1 | 0% | 862 | 991 | +15% | 0 | 0 | — |
case-24 | pass→pass | 15,512 | 2,570 | -83% | 1 | 1 | 0% | 1,301 | 1,002 | -23% | 0 | 0 | — |
case-25 | pass→pass | 11,126 | 20,018 | +80% | 1 | 1 | 0% | 1,015 | 2,449 | +141% | 0 | 0 | — |
case-26 | pass→pass | 14,992 | 13,672 | -9% | 1 | 1 | 0% | 1,762 | 2,207 | +25% | 0 | 0 | — |
case-27 | fail→fail | 25,591 | 18,997 | -26% | 1 | 1 | 0% | 1,088 | 2,113 | +94% | 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. 27 cases were attempted, and 25 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 +30 percentage points is the difference between those two pass rates over the 25 comparable cases. 1 case got worse with the skill loaded, and it is 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.