Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Report the current state of the context window and recommend compression or cleanup actions. Activate when user mentions "context health", "context window", "how much context", "context full", "running out of context", "compress context".
.claude/skills/bilal140202-doncheli-context-health/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 41% | 0% |
## Context Health — 2026-03-28T16:00Z
### Usage Estimate
- Current session: ~85K tokens
- Model limit: 200K tokens
- Usage: 42% ✅ (healthy)
### Largest Items
1. src/payments/processor.ts full read — ~8K tokens (can evict — already analyzed)
2. Test run output from 14:32 — ~12K tokens (can summarize — report saved)
3. Conversation history — ~18K tokens (normal)
### Recommendations
- No immediate action required
- If you add 3+ more large files, consider /dc:destilar to compress session notes
- Next heavy task: use a subagent to isolate output
### Warning Threshold
Will warn again at 70% usage (~140K tokens)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 8,813 | 4,373 | -50% | 1 | 1 | 0% | 1,535 | 1,296 | -16% | 0 | 0 | — |
case-02 | fail→pass | 7,830 | 3,151 | -60% | 1 | 1 | 0% | 1,330 | 987 | -26% | 0 | 0 | — |
case-03 | fail→pass | 9,090 | 5,563 | -39% | 1 | 1 | 0% | 1,478 | 1,349 | -9% | 0 | 0 | — |
case-04 | fail→pass | 10,420 | 6,875 | -34% | 1 | 1 | 0% | 1,797 | 1,699 | -5% | 0 | 0 | — |
case-05 | pass→pass | 5,641 | 4,293 | -24% | 1 | 1 | 0% | 949 | 1,231 | +30% | 0 | 0 | — |
case-06 | pass→pass | 8,124 | 3,338 | -59% | 1 | 1 | 0% | 1,485 | 1,062 | -28% | 0 | 0 | — |
case-07 | pass→pass | 6,572 | 3,463 | -47% | 1 | 1 | 0% | 1,073 | 1,143 | +7% | 0 | 0 | — |
case-08 | fail→pass | 6,006 | 5,605 | -7% | 1 | 1 | 0% | 980 | 1,383 | +41% | 0 | 0 | — |
case-09 | fail→pass | 12,397 | 4,686 | -62% | 1 | 1 | 0% | 2,246 | 1,179 | -48% | 0 | 0 | — |
case-10 | fail→pass | 7,608 | 3,932 | -48% | 1 | 1 | 0% | 1,312 | 1,127 | -14% | 0 | 0 | — |
case-11 | fail→pass | 10,496 | 2,559 | -76% | 1 | 1 | 0% | 1,779 | 908 | -49% | 0 | 0 | — |
case-12 | fail→pass | 8,680 | 1,687 | -81% | 1 | 1 | 0% | 1,685 | 654 | -61% | 0 | 0 | — |
case-13 | fail→pass | 13,801 | 3,727 | -73% | 1 | 1 | 0% | 2,368 | 1,128 | -52% | 0 | 0 | — |
case-14 | fail→pass | 11,347 | 1,587 | -86% | 1 | 1 | 0% | 841 | 702 | -17% | 0 | 0 | — |
case-15 | pass→pass | 9,462 | 3,578 | -62% | 1 | 1 | 0% | 1,687 | 1,063 | -37% | 0 | 0 | — |
case-16 | fail→fail | 3,698 | 4,115 | +11% | 1 | 1 | 0% | 841 | 1,309 | +56% | 0 | 0 | — |
case-17 | fail→pass | 6,507 | 2,582 | -60% | 1 | 1 | 0% | 1,162 | 689 | -41% | 0 | 0 | — |
case-18 | pass→pass | 5,183 | 8,794 | +70% | 1 | 1 | 0% | 1,115 | 2,283 | +105% | 0 | 0 | — |
case-19 | pass→pass | 6,836 | 9,581 | +40% | 1 | 1 | 0% | 1,420 | 2,232 | +57% | 0 | 0 | — |
case-20 | pass→pass | 10,436 | 8,612 | -17% | 1 | 1 | 0% | 1,919 | 2,112 | +10% | 0 | 0 | — |
case-21 | pass→pass | 9,524 | 6,261 | -34% | 1 | 1 | 0% | 1,708 | 1,496 | -12% | 0 | 0 | — |
case-22 | fail→pass | 8,232 | 1,775 | -78% | 1 | 1 | 0% | 1,347 | 673 | -50% | 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 21 counted toward the lift figure. The other 1 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 +59 percentage points is the difference between those two pass rates over the 21 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.