radiology-method-evaluation
Audit parameter provenance, metric fit, fair comparisons and robustness; not execution or paper review.
Install / Use
npx skills add huang-sir1/radiology-skills --skill radiology-method-evaluationInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Content & MediaSupported Platforms
Our assessment of radiology-method-evaluation
radiology-method-evaluation scores 82/100 on our quality scale, 969th of 1,153 Content & Media skills we index.
Its SKILL.md is 11 KB long, well organised into 9 sections and no code examples: a thorough specification that gives an agent plenty to work with.
With 1,785 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 2 months ago, so radiology-method-evaluation is actively maintained.
- Our last check on 2026-11-07 found the source still online.
- No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
- Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-10-01. Automated pattern scan on 2026-10-01. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
radiology-method-evaluation compared with similar skills
All 4 of these similar skills score higher than radiology-method-evaluation; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| radiology-method-evaluation (this skill)by huang-sir1 | 82 | 1.8k | 56d ago | SKILL.md |
| siyuanby siyuan-note | 100 | 47.3k | today | MCP Server |
| x402by browser-use | 100 | 121.0k | 3d ago | SKILL.md |
| algorithmic-artby anthropics | 100 | 177.9k | 55d ago | SKILL.md |
| pptxby anthropics | 100 | 177.9k | 55d ago | SKILL.md |
Frequently asked questions
- How do I install radiology-method-evaluation?
- Run
npx skills add huang-sir1/radiology-skills --skill radiology-method-evaluation. The install tabs above show the steps for each supported agent. - Which AI agents does radiology-method-evaluation work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is radiology-method-evaluation safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is radiology-method-evaluation still maintained?
- The repository was last updated about 2 months ago, so radiology-method-evaluation is actively maintained.
Skill content
View source on GitHubname: radiology-method-evaluation description: "Audit parameter provenance, metric fit, fair comparisons and robustness; not execution or paper review."
Parameter and methodology evaluation
Use this skill to decide whether a parameter choice, evaluation metric, or analytical method is scientifically defensible and whether the available evaluation is sufficient for the stated claim. It is an independent entry point and a cross-cutting reviewer/mentor module, not a sub-skill of statistics, radiogenomics, writing, or prereview.
Core stance
- Treat parameters/settings, evaluation metrics/readouts, and methodology as three separate objects. Do not hide one behind another.
- Use exact evidence anchors. Report whether evidence is
PLANNED,AUTHOR_REPORTED,PARTLY_VERIFIED,VERIFIED, orNOT_ASSESSABLE. - Return
PASS,CONDITIONAL, orSTOPper decision-bearing object and claim. Do not average a fatal defect into a composite quality score. - Be constructive: identify the criterion, minimum feasible repair, stronger option and its cost, closure evidence, and the claim that survives if the repair cannot be completed.
- Review is read-only by default. Do not alter code, data, configurations, or manuscript files unless the user explicitly asks for implementation or editing.
- Never invent a parameter range, run, result, effect, uncertainty interval, software version, benchmark, ablation, or validation.
Choose the entry mode
| Mode | Use when | Required output emphasis |
|---|---|---|
| plan | the evaluation has not been run | decision-bearing parameters, defensible alternatives/ranges, fair comparison design, controls, freeze point and analysis handoff |
| audit | Methods, configurations, code, logs or result artifacts exist | provenance, selection timing, leakage, metric fit, method assumptions, fairness, evidence gaps and atomic repairs |
| interpret | real evaluation results are supplied | direction/magnitude, uncertainty, stability/failure boundary, heterogeneity, claim consequence and residual limit |
| writing-package | the science is fixed and must be placed in a paper | exact Methods/Results/Figure/Supplement/Discussion jobs without implying unrun analyses |
Also declare focus: parameters | evaluation-metrics | methodology | combined and the active
scope/modalities. A request may enter with an incomplete manuscript or a single configuration file;
do not require a prior pipeline task.
Minimal intake
Record what is available and mark the rest AUTHOR_INPUT_NEEDED:
- study question, estimand or intended use and affected Claim IDs;
- scope and modalities; cohort/specimen hierarchy and true independent unit;
- frozen method/model/pipeline version and data/split/test-access state;
- parameter values, defaults, candidate ranges/levels, units and rationale sources;
- when and on which data each setting, threshold, model or metric was selected;
- comparators, controls, ablations, evaluation metrics, tuning/compute budget and hardware;
- supplied results, uncertainty, run/seed structure, failures and exact artifact locations.
Required workflow
- Open the canonical assessment reference.
- Reconstruct the chain:
question/estimand -> unit -> measurement -> preprocessing -> representation -> selection -> method/model -> metric -> validation -> claim. - Build separate methodology-fit, parameter, metric/readout, and sensitivity/robustness tables.
- Activate only evaluations that can change a decision or claim. Do not demand every possible grid, ablation, or alternative method.
- For each finding, bind the exact evidence anchor, criterion, affected claim, minimum repair, optional stronger route/cost, closure evidence, and manuscript placement.
- If actual comparison outputs need statistical inference, hand off the paired predictions or
estimates, independent unit, resampling hierarchy and multiplicity family to
radiology-stats. - Use the reusable assessment matrix when a durable artifact is useful.
- When the team wants to reuse institution- or platform-specific experience, open the local experience registry. Transfer only rows whose platform, specimen, data state, endpoint, unit and intended claim are compatible; preserve failed and null experience as well as successful settings.
Scientific red lines
- Tuning selects a configuration; sensitivity analysis evaluates a locked conclusion. Do not present the tuning trace as independent robustness evidence.
- The protected test set cannot select parameters, methods, thresholds, features, stopping rules, favourable seeds, figures or claims.
- Fair method comparison holds cases, splits, preprocessing, endpoint, metric and—where relevant— tuning budget, pretraining information, implementation and hardware conditions comparable.
- Folds, seeds, patches, slices, lesions, cells, spots, tiles and model runs are not new patients or donors. Preserve the actual sampling and resampling hierarchy.
- If changing a setting changes the estimand, inclusion population, spatial scale, cell identity, biological object or decision threshold, treat it as a distinct analysis question—not a routine sensitivity check.
- Perturbation dose, time, guide design/number, assignment and target-engagement criteria are experimental-design factors. Do not optimize them against the desired phenotype as if they were ordinary machine-learning hyperparameters.
- An ablation supports the component's conditional contribution in the evaluated pipeline and data; it does not prove universal necessity, biological mechanism or causality.
- Stability is not correctness. Simulation, generated molecular layers and synthetic perturbations are not external or biological validation.
- A package default, popular paper or high-impact venue is not by itself a scientific rationale.
Domain routing
Use this skill as the evaluation owner; open the relevant internal domain reference only for the meaning, plausible alternatives and operational constraints of the active parameter or method.
| Domain | Domain execution/source owner | Typical decision-bearing evaluation |
|---|---|---|
| image formation/quantitation | radiology-acquisition-qc | series/phase/sequence acceptance, DICOM quantitative transforms, acquisition/reconstruction, contrast/tracer timing, dose, artifacts, phantom/test-retest, site qualification and protocol drift |
| radiomics/imaging analysis | radiology-radiomics, radiology-annotation | resampling/interpolation, binning, segmentation perturbation, stability cutoff, harmonisation, feature selection and threshold freeze after the measurement passport |
| deep imaging | radiology-deep-learning | input scale/patch, architecture/pretraining, augmentation, loss/optimizer/schedule, search budget, seeds, checkpointing, ablation, OOD and efficiency |
| bulk RNA | execution: radiology-transcriptomics-analysis; meaning: radiology-radiogenomics bulk playbook | filtering, normalization, design/contrast, batch/composition, model family, FDR/effect thresholds, gene-set universe |
| sc/snRNA | execution: radiology-transcriptomics-analysis; meaning: radiology-radiogenomics single-cell playbook | QC, ambient/doublet handling, normalization/HVG, integration, PCs/kNN/resolution, annotation, pseudobulk unit, trajectory root |
| spatial | execution: radiology-transcriptomics-analysis; meaning: radiology-radiogenomics spatial playbook | segmentation, coordinate error, deconvolution, smoothing, domain resolution/seed, radius/kernel/k, spatial null and patient aggregation |
| pathology | radiology-radiogenomics pathology playbook | stain/scoring, magnification/tile, colour normalization, segmentation, phenotype cutoff, ROI sampling and patient aggregation |
| multi-omics | radiology-radiogenomics integration playbooks | matched intersection, missing-modality policy, scaling, factor/latent dimension, modality weights, alignment objective, unimodal ablation |
| perturbation | design/audit source: radiology-experiment-design; mechanism meaning: radiology-radiogenomics perturbation playbook | assignment, efficiency, off-target, dose-time, target engagement, toxicity, control/rescue and multiplicity |
| imaging-mechanism | radiology-radiogenomics bridge playbook | all active modality settings plus patient-lesion-region-block-section-cell/time mapping, bridge sensitivity and discordance |
Output contract
Return the shortest complete set needed for the request, in this order:
Evaluation scope and claim— mode, focus, scope, active modalities, independent unit, materials and evidence boundary.Methodology-fit matrix— question/estimand, data layer/unit, assumptions, comparator/control, validation and verdict.Parameter ledger— role, value/range/unit, rationale, selection data/timing, search budget, freeze state, sensitivity, provenance and claim consequence.Evaluation-metric/readout matrix— task fit, aggregation, threshold/prevalence dependence, uncertainty, tuning-objective overlap and blind spot.Benchmark/ablation/control matrixandSensitivity/robustness results or plan— matched conditions, execution state, failure boundary and negative results.Strengths and atomic findings— evidence anchor, criterion, severity, minimum repair, stronger option/cost and closure evidence.Statistical handoff— only the comparisons requiring estimation, intervals, tests or multiplicity control.Writing placement— Methods, Results, Figure/Table, Supplement/config/code, Discussion and Abstract/title constraints.Claim verdict and next action—PASS / CONDITIONAL / STOP, surviving claim ceiling, the single next decision-bearing action andAUTHOR_INPUT_NEEDEDfields.
Handoffs and non-ownership
- Configure or run a radiomics/DL pipeline → its imaging execution skill; configure, run, audit or
reproduce bulk/sc/snRNA/spatial workflows →
radiology-transcriptomics-analysis. - Design functional IHC/mIF, tissue, cell, organoid, animal, perturbation, target-engagement or
rescue evidence →
radiology-experiment-design; this skill may later audit its method choices. - Estimate paired differences, CIs, cluster bootstrap, run variability or multiplicity →
radiology-statsafter this skill defines the comparison and hierarchy. - Audit all dimensions of a full manuscript →
radiology-prereview; it consumes these findings without replacing the criterion or closure evidence. - Draft prose from fixed evidence →
radiology-writing; sentence-level polish →radiology-polishing. - Turn reviewer comments into a response/verification ledger →
radiology-response. - Audit journal upload files and format rules →
radiology-submission.
Do not silently broaden a focused parameter/method question into a full manuscript review or an unrequested execution project.
Related Skills
siyuan
47.3kAn open-source, privacy-first, self-hosted knowledge workspace where humans and AI agents work together 开源、隐私优先、自托管的知识工作空间,让人与智能体在此协作
x402
121.0kSet up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key.
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
pptx
177.9kUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an em…
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
