Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prune bloated session with a prescription. Removes progress ticks, stale reads, duplicate content, and more.
.claude/skills/ruya-ai-treat/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -41% | 0% |
Apply a pruning prescription to the current session. Default is standard if no argument given.
When to use treat vs reload: prefer the reload skill — it does the same prune plus auto-resume in one step (/exit → new terminal opens with the pruned session). Use treat only when the user explicitly wants to stay in the current session, work in a multi-pane setup, or resume manually.
bash cozempic current --diagnose
bash cozempic treat current -rx $ARGUMENTS If no argument was provided, use standard: bash cozempic treat current -rx standard
Tokens: line). Always surface both byte and token savings.bash cozempic treat current -rx $ARGUMENTS --execute
claude --resume. (Tip: next time, the reload skill does this in one step — /exit and a fresh terminal opens automatically.)"| Rx | Strategies | Typical Savings | |----|-----------|----------------| | gentle | progress-collapse, file-history-dedup, metadata-strip | 40-55% | | standard | gentle + thinking-blocks, tool-output-trim, stale-reads, system-reminder-dedup | 50-70% | | aggressive | standard + error-retry-collapse, document-dedup, mega-block-trim, envelope-strip | 70-95% |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,655 | 9,024 | +36% | 1 | 1 | 0% | 1,034 | 779 | -25% | 0 | 0 | — |
case-02 | fail→fail | 7,656 | 4,591 | -40% | 1 | 1 | 0% | 1,086 | 814 | -25% | 0 | 0 | — |
case-03 | fail→fail | 5,899 | 2,580 | -56% | 1 | 1 | 0% | 920 | 846 | -8% | 0 | 0 | — |
case-04 | fail→fail | 7,416 | 3,776 | -49% | 1 | 1 | 0% | 1,144 | 709 | -38% | 0 | 0 | — |
case-05 | fail→fail | 7,253 | 3,794 | -48% | 1 | 1 | 0% | 1,236 | 801 | -35% | 0 | 0 | — |
case-06 | fail→fail | 11,868 | 4,225 | -64% | 1 | 1 | 0% | 1,824 | 756 | -59% | 0 | 0 | — |
case-07 | fail→fail | 3,625 | 2,738 | -24% | 1 | 1 | 0% | 636 | 769 | +21% | 0 | 0 | — |
case-08 | fail→fail | 6,362 | 5,730 | -10% | 1 | 1 | 0% | 1,054 | 815 | -23% | 0 | 0 | — |
case-09 | fail→fail | 2,423 | 7,117 | +194% | 1 | 1 | 0% | 401 | 967 | +141% | 0 | 0 | — |
case-10 | fail→fail | 5,878 | 3,979 | -32% | 1 | 1 | 0% | 912 | 704 | -23% | 0 | 0 | — |
case-11 | fail→fail | 7,312 | 4,437 | -39% | 1 | 1 | 0% | 1,139 | 761 | -33% | 0 | 0 | — |
case-12 | fail→pass | 10,073 | 2,186 | -78% | 1 | 1 | 0% | 1,623 | 910 | -44% | 0 | 0 | — |
case-13 | fail→pass | 10,605 | 1,635 | -85% | 1 | 1 | 0% | 1,833 | 770 | -58% | 0 | 0 | — |
case-14 | pass→pass | 6,824 | 4,107 | -40% | 1 | 1 | 0% | 1,043 | 1,173 | +12% | 0 | 0 | — |
case-15 | fail→pass | 10,829 | 2,561 | -76% | 1 | 1 | 0% | 1,755 | 898 | -49% | 0 | 0 | — |
case-16 | pass→pass | 11,005 | 2,749 | -75% | 1 | 1 | 0% | 1,635 | 859 | -47% | 0 | 0 | — |
case-17 | fail→fail | 4,276 | 7,702 | +80% | 1 | 1 | 0% | 685 | 986 | +44% | 0 | 0 | — |
case-18 | fail→pass | 7,128 | 1,424 | -80% | 1 | 1 | 0% | 987 | 784 | -21% | 0 | 0 | — |
case-19 | fail→pass | 7,941 | 1,572 | -80% | 1 | 1 | 0% | 1,208 | 709 | -41% | 0 | 0 | — |
case-20 | fail→fail | 6,896 | 5,619 | -19% | 1 | 1 | 0% | 1,248 | 866 | -31% | 0 | 0 | — |
case-21 | pass→fail | 3,419 | 5,051 | +48% | 1 | 1 | 0% | 612 | 759 | +24% | 0 | 0 | — |
case-22 | pass→pass | 8,720 | 3,851 | -56% | 1 | 1 | 0% | 1,499 | 907 | -39% | 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 11 counted toward the lift figure. The other 11 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 11 comparable cases. 2 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.