- Home
- Skills
- Technical Documentation
- methodology-designer
Works with the AI Agents you already use
Methodology Designer
Design a rigorous, reproducible methodology, with baselines, metrics, and ablations reviewers cannot reject.
$8
methodology-designer
Example session with this skill installed
Use methodology-designer to design an experimental methodology for comparing three deep learning models on skin-lesion classification.
- Read your context and instructions
- Compiled the methodology-designer
Designed a reproducible methodology for a three-model comparison (ResNet-50, EfficientNet-B4, ViT-B/16) on skin-lesion classification, with explicit baselines, metrics, ablations, and a threat-to-validity plan.
Problem definition
Multi-class classification of dermoscopic images into 7 lesion categories (HAM10000 classes). Research question: does a vision transformer outperform CNNs under class imbalance and limited labeled data?
Experimental protocol
| Component | Choice | Justification |
|---|---|---|
| Dataset | HAM10000 + ISIC 2019 | Public, multi-class, standard benchmark |
| Split | Stratified 70/15/15 | Preserves rare-class proportions |
| Preprocessing | 224×224, color constancy | Matches pretrained backbones |
| Baselines | ResNet-50, EfficientNet-B4 | Strong CNN controls |
| Proposed | ViT-B/16 (ImageNet-21k pretrained) | Tests transformer claim |
| Metrics | Balanced accuracy, macro-F1, AUC | Handles class imbalance |
| Statistical test | McNemar + 5×2 cross-validation | Justifies significance |
Ablation plan
- Pretraining source (ImageNet vs. self-supervised DINO).
- Input resolution (224 vs. 384).
- Data augmentation (none, light, heavy).
Reproducibility
- Seed all randomness; release configs, optimizer settings, and the augmentation pipeline.
- Report hardware and training time; use a fixed environment (Docker).
Next steps
- Run
figure-designerto turn the method into a methodology or pipeline figure. - Run
paper-outlineto map these into an IMRaD structure. - Run
reviewerto stress-test the design before running experiments.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
A weak methodology is the number one reason reviewers reject technically strong work. Researchers often describe what they did without justifying why each design choice was made, which leaves reviewers unable to trust the results. This skill designs a rigorous, reproducible methodology that stands up to scrutiny.
What it does
- Structures the method description and formalizes the problem definition.
- Designs the experimental protocol and selects appropriate baselines and metrics.
- Justifies every design choice with the alternatives considered and the expected effect.
- Plans ablation studies, sensitivity analysis, and statistical testing.
- Produces a reproducibility checklist with enough implementation detail to rerun the work.
Why this beats prompting it yourself
Generic prompts return a vague methods section that describes the approach but never justifies it. This skill follows a five-component framework, covering problem formalization, architecture, design justification, experimental protocol, and reproducibility, so every choice is tied to evidence and every experiment answers a specific research question.
Use cases
- Writing the methodology section of a paper or PhD proposal.
- Designing experiments before running them, instead of reverse-engineering them afterward.
- Selecting honest baselines and metrics that reviewers cannot dismiss.
- Planning ablation studies that prove which components actually matter.
- Reviewing an existing methodology for weaknesses before submission.
Known limitations
The design is only as good as the information you provide; it needs specific inputs about the problem scope, the proposed approach, the key innovation, and the available resources to be effective. It does not perform actual code execution or statistical computation, and it does not run the experiments; it produces a specification, so following the reproducibility checklist is still your responsibility.
Important
It designs the methodology and experimental protocol; it does not run the experiments or produce results for you. Pair it with paper-outline to map the design into a manuscript structure.
Why buy it
Use it when writing a methodology section, designing experiments, or planning a project's technical approach. It turns a fragile, hand-wavy methods section into a rigorous, reproducible plan.
Requires user-provided problem details, key innovations, and baseline methods. Does not execute code or run actual experiments.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Permissions
No special permissions declared or detected