Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generates executable Python protocols for Opentrons OT-2 and Flex robots from natural language descriptions.
.claude/skills/opentrons-protocol-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 161% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | 3% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -15% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 24% | 0% |
<!--
#
#
-->
The Opentrons Protocol Agent bridges the gap between experimental design and physical execution by translating instructions into Opentrons Python API scripts.
.py files for OT-2/Flex.opentrons_protocol_template.py or internal logic to draft the script.User: "Create a protocol for 1:2 serial dilution in a 96-well plate."
Agent Action: (Uses an LLM to fill in the template or generate code)
bash# Agent writes the file to disk python3 -m opentrons.simulate generated_protocol.py
<!-- 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-01 | pass→pass | 32,599 | 12,589 | -61% | 1 | 1 | 0% | 3,114 | 3,221 | +3% | 0 | 0 | — |
case-02 | pass→pass | 37,136 | 15,437 | -58% | 1 | 1 | 0% | 4,396 | 3,746 | -15% | 0 | 0 | — |
case-03 | pass→pass | 7,088 | 6,645 | -6% | 1 | 1 | 0% | 1,444 | 1,789 | +24% | 0 | 0 | — |
case-04 | pass→pass | 5,753 | 6,287 | +9% | 1 | 1 | 0% | 1,162 | 1,750 | +51% | 0 | 0 | — |
case-05 | pass→pass | 15,210 | 16,506 | +9% | 1 | 1 | 0% | 2,756 | 3,504 | +27% | 0 | 0 | — |
case-06 | pass→pass | 6,101 | 3,868 | -37% | 1 | 1 | 0% | 1,133 | 1,229 | +8% | 0 | 0 | — |
case-07 | fail→fail | 10,307 | 7,261 | -30% | 1 | 1 | 0% | 2,179 | 1,883 | -14% | 0 | 0 | — |
case-08 | pass→pass | 12,817 | 9,230 | -28% | 1 | 1 | 0% | 2,647 | 2,230 | -16% | 0 | 0 | — |
case-09 | pass→pass | 6,189 | 7,358 | +19% | 1 | 1 | 0% | 1,198 | 1,877 | +57% | 0 | 0 | — |
case-10 | pass→pass | 7,248 | 6,104 | -16% | 1 | 1 | 0% | 1,417 | 1,731 | +22% | 0 | 0 | — |
case-11 | pass→pass | 7,413 | 6,030 | -19% | 1 | 1 | 0% | 1,416 | 1,649 | +16% | 0 | 0 | — |
case-12 | pass→pass | 9,206 | 7,192 | -22% | 1 | 1 | 0% | 1,805 | 2,013 | +12% | 0 | 0 | — |
case-13 | pass→pass | 11,179 | 7,114 | -36% | 1 | 1 | 0% | 2,184 | 1,903 | -13% | 0 | 0 | — |
case-14 | fail→pass | 17,570 | 7,786 | -56% | 1 | 1 | 0% | 1,392 | 1,932 | +39% | 0 | 0 | — |
case-15 | pass→pass | 10,900 | 8,048 | -26% | 1 | 1 | 0% | 2,007 | 2,070 | +3% | 0 | 0 | — |
case-16 | pass→pass | 11,206 | 9,422 | -16% | 1 | 1 | 0% | 2,163 | 2,265 | +5% | 0 | 0 | — |
case-17 | fail→pass | 9,183 | 3,384 | -63% | 1 | 1 | 0% | 379 | 990 | +161% | 0 | 0 | — |
case-18 | fail→fail | 14,278 | 8,516 | -40% | 1 | 1 | 0% | 2,730 | 2,148 | -21% | 0 | 0 | — |
case-19 | pass→pass | 8,389 | 7,019 | -16% | 1 | 1 | 0% | 1,640 | 1,787 | +9% | 0 | 0 | — |
case-20 | pass→pass | 7,340 | 10,604 | +44% | 1 | 1 | 0% | 1,418 | 2,862 | +102% | 0 | 0 | — |
case-21 | pass→pass | 10,301 | 7,919 | -23% | 1 | 1 | 0% | 2,041 | 1,974 | -3% | 0 | 0 | — |
case-22 | pass→pass | 11,584 | 11,572 | -0% | 1 | 1 | 0% | 2,254 | 2,588 | +15% | 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 +9 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/26/2026 | +9% |
Other measured skills in the registry, with their headline benchmark lift.