Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →sweetcornna/mathodology-agent-pipeline
Use when planning a modeling solution, selecting the next useful step or briefing a specialist.
Passed 11 of 24 test cases with the skill, and 9 of 24 without it.
and 2 more in the full report
4 fixed7 passed either way11 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 4 test cases the skill fixed. Same model, same prompt, with and without the skill.
We are starting a fresh mathematical modeling challenge on optimizing emergency drone delivery routes for medical supplies. What is a compact, effective prompt directive to launch the agent workflow?
| Checked | Without skill | With skill |
|---|---|---|
| Directs defining assumptions and building a useful baseline model. | Passed | Passed |
| Directs selecting purposeful figures using mathodology-figure-presets and asking once about image2 availability. | Missed | Passed |
The instructions your model reads when the skill loads.
Use the following questions in whatever order the task needs. They are prompts for reasoning, not mandatory stages or files.
What decision is the reader trying to make? What is given, unknown or required? Which mechanisms must the solution represent? Check the actual contest rules when applicable, including deadline, page limits and AI-use requirements.
Start with a useful baseline. Define variables, units, assumptions, constraints and the objective. Compare plausible alternatives when there is a real choice; do not invent extra models to meet a quota. Explain why the added complexity changes the answer. Check identifiability, data requirements and limiting cases.
Which observation could disprove the model? Can a simpler baseline perform as well? Test influential assumptions and plausible adverse scenarios. Separate parameter uncertainty, observation noise and structural uncertainty. Match the paper's claims to the implemented mathematics and the data actually used.
Answer the problem's questions with interpretable quantities and limitations. Choose figures from figure presets, including the once-per-task image2 question. Build the explanation around the results, not the history of experiments. Review with review questions.
When delegation is useful and available, give a specialist a bounded question, relevant data, current assumptions and a concrete output. Agree file ownership for concurrent editing. Ask for ordinary prose: finding, reasoning, artifact paths and unresolved uncertainty. The lead integrates the answer and resolves conflicting evidence; it does not collect points or gate every intermediate step.
test cases passed, vs 9 / 24 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.