Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →AI-powered analysis for predicting optimal immune checkpoint inhibitor combinations based on tumor microenvironment, biomarkers, and molecular profiling.
.claude/skills/immune-checkpoint-combination-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-16 | ✓→✓ | ＝ Same ✓ | 63% | 0% |
| case-22 | ✓→✓ | ＝ Same ✓ | 26% | 0% |
<!--
#
#
-->
The Immune Checkpoint Combination Agent analyzes tumor molecular profiles to predict optimal immune checkpoint inhibitor (ICI) combinations. It integrates TME characterization, checkpoint expression, resistance mechanisms, and clinical evidence for rational immunotherapy combination design.
| Target | Approved Agents | Mechanism | Combination Rationale | |--------|-----------------|-----------|----------------------| | PD-1 | Pembrolizumab, Nivolumab | Block T-cell inhibition | Backbone therapy | | PD-L1 | Atezolizumab, Durvalumab | Block tumor immune evasion | Alternative backbone | | CTLA-4 | Ipilimumab, Tremelimumab | Enhance T-cell priming | Non-redundant to PD-1 | | LAG-3 | Relatlimab | Block exhausted T-cells | PD-1 refractory | | TIGIT | Tiragolumab | Block NK/T suppression | NK cell engagement | | TIM-3 | Multiple in trials | Terminal exhaustion | Highly exhausted TME |
User: "Recommend optimal checkpoint inhibitor combination for this melanoma patient based on their tumor profile."
Agent Action:
bashpython3 Skills/Immunology_Vaccines/Immune_Checkpoint_Combination_Agent/ici_combination.py \ --rnaseq tumor_expression.tsv \ --ihc pd-l1_tps_60.json \ --mutations tumor_mutations.maf \ --tmb 12.5 \ --msi stable \ --tumor_type melanoma \ --prior_treatment pembrolizumab \ --output ici_recommendations.json
Inflamed ("Hot") Tumors:
Excluded Tumors:
Desert ("Cold") Tumors:
| Mechanism | Biomarkers | Combination Strategy | |-----------|------------|---------------------| | Alternative checkpoints | LAG-3+, TIGIT+, TIM-3+ | Add second checkpoint | | WNT/β-catenin | CTNNB1 mutations | Poor ICI candidate | | IFN signaling loss | JAK1/2, B2M mutations | Limited benefit | | MHC loss | HLA-A/B/C loss | NK-engaging therapies | | T-cell exclusion | TGF-β high | TGF-β inhibitor combination |
Response Prediction:
Synergy Prediction:
| Combination | Indication | Key Trial | Benefit | |-------------|------------|-----------|---------| | Nivo + Ipi | Melanoma | CheckMate-067 | OS improvement | | Nivo + Rela | Melanoma | RELATIVITY-047 | PFS improvement | | Atezo + Tira | NSCLC | CITYSCAPE | PFS improvement (PD-L1 high) | | Durva + Treme | HCC | HIMALAYA | OS improvement |
AI Group - Biomedical AI Platform
<!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-16 | pass→pass | 11,808 | 11,627 | -2% | 1 | 1 | 0% | 2,121 | 3,451 | +63% | 0 | 0 | — |
case-22 | pass→pass | 50,028 | 24,440 | -51% | 1 | 1 | 0% | 4,606 | 5,814 | +26% | 0 | 0 | — |
case-01 | fail→fail | 22,617 | 17,093 | -24% | 1 | 1 | 0% | 4,109 | 4,437 | +8% | 0 | 0 | — |
case-02 | fail→pass | 21,972 | 2,981 | -86% | 1 | 1 | 0% | 1,345 | 2,096 | +56% | 0 | 0 | — |
case-03 | fail→pass | 14,738 | 10,980 | -25% | 1 | 1 | 0% | 2,678 | 3,328 | +24% | 0 | 0 | — |
case-04 | pass→pass | 10,243 | 9,612 | -6% | 1 | 1 | 0% | 1,784 | 3,097 | +74% | 0 | 0 | — |
case-05 | pass→pass | 10,483 | 8,998 | -14% | 1 | 1 | 0% | 1,941 | 3,072 | +58% | 0 | 0 | — |
case-06 | pass→pass | 12,440 | 11,193 | -10% | 1 | 1 | 0% | 2,356 | 3,465 | +47% | 0 | 0 | — |
case-07 | pass→pass | 9,007 | 9,576 | +6% | 1 | 1 | 0% | 1,741 | 3,231 | +86% | 0 | 0 | — |
case-08 | pass→pass | 7,518 | 10,388 | +38% | 1 | 1 | 0% | 1,338 | 3,359 | +151% | 0 | 0 | — |
case-09 | pass→pass | 15,139 | 10,802 | -29% | 1 | 1 | 0% | 2,222 | 3,237 | +46% | 0 | 0 | — |
case-14 | pass→pass | 14,915 | 13,676 | -8% | 1 | 1 | 0% | 2,774 | 3,999 | +44% | 0 | 0 | — |
case-15 | pass→pass | 17,000 | 15,203 | -11% | 1 | 1 | 0% | 3,043 | 4,188 | +38% | 0 | 0 | — |
case-10 | pass→pass | 5,532 | 6,224 | +13% | 1 | 1 | 0% | 1,042 | 2,655 | +155% | 0 | 0 | — |
case-11 | pass→pass | 4,722 | 6,378 | +35% | 1 | 1 | 0% | 917 | 2,666 | +191% | 0 | 0 | — |
case-12 | pass→pass | 8,403 | 3,998 | -52% | 1 | 1 | 0% | 1,791 | 2,115 | +18% | 0 | 0 | — |
case-13 | pass→pass | 7,433 | 5,499 | -26% | 1 | 1 | 0% | 769 | 2,414 | +214% | 0 | 0 | — |
case-17 | fail→pass | 15,292 | 2,376 | -84% | 1 | 1 | 0% | 1,338 | 1,990 | +49% | 0 | 0 | — |
case-18 | pass→pass | 8,759 | 9,468 | +8% | 1 | 1 | 0% | 1,631 | 3,140 | +93% | 0 | 0 | — |
case-19 | pass→pass | 13,356 | 17,583 | +32% | 1 | 1 | 0% | 2,145 | 4,426 | +106% | 0 | 0 | — |
case-20 | pass→pass | 25,334 | 14,815 | -42% | 1 | 1 | 0% | 3,725 | 4,147 | +11% | 0 | 0 | — |
case-21 | pass→pass | 12,329 | 12,675 | +3% | 1 | 1 | 0% | 2,840 | 4,091 | +44% | 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. The headline lift of +14 percentage points is the difference between those two pass rates over the 22 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/26/2026 | +18% |
Other measured skills in the registry, with their headline benchmark lift.