radiology-experiment-design
Design/audit tissue, cell, organoid, animal, perturbation and rescue validation; not physical SOP execution.
Install / Use
npx skills add huang-sir1/radiology-skills --skill radiology-experiment-designInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Development & EngineeringSupported Platforms
Our assessment of radiology-experiment-design
radiology-experiment-design scores 84/100 on our quality scale, 2888th of 4,700 Development & Engineering skills we index.
Its SKILL.md is 14 KB long, well organised into 10 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-experiment-design is actively maintained.
- Our last check on 2026-11-05 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-experiment-design compared with similar skills
All 4 of these similar skills score higher than radiology-experiment-design; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| radiology-experiment-design (this skill)by huang-sir1 | 84 | 1.8k | 56d ago | SKILL.md |
| taste-skillby Leonxlnx | 100 | 116.8k | 8d ago | CLAUDE.md |
| ai-job-searchby ai-job-search-dev | 100 | 72.1k | today | CLAUDE.md |
| claude-howtoby luongnv89 | 100 | 47.3k | 2d ago | CLAUDE.md |
| Claude-Code-Game-Studiosby Donchitos | 100 | 39.7k | 7d ago | CLAUDE.md |
Frequently asked questions
- How do I install radiology-experiment-design?
- Run
npx skills add huang-sir1/radiology-skills --skill radiology-experiment-design. The install tabs above show the steps for each supported agent. - Which AI agents does radiology-experiment-design 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-experiment-design 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-experiment-design still maintained?
- The repository was last updated about 2 months ago, so radiology-experiment-design is actively maintained.
Skill content
View source on GitHubname: radiology-experiment-design description: "Design/audit tissue, cell, organoid, animal, perturbation and rescue validation; not physical SOP execution."
Mechanism experiment design and review
Use this skill as an independent entry point for turning an imaging-mechanism hypothesis into a feasible, falsifiable and auditable experimental programme. It can also review an existing plan or result package and teach the author how to improve it. It does not claim that an experiment was run and never substitutes generic advice for a laboratory's validated SOP.
Core stance
- Start from a decision-bearing claim and at least one competing explanation, not from a preferred assay. Ask what observation would distinguish the explanations.
- Preserve an evidence ladder: observational alignment, orthogonal measurement, perturbation, target engagement, phenotype, rescue/reversal and replication are different evidential jobs.
- A perturbation-associated change supports a causal interpretation only to the extent justified by assignment, specificity, target engagement, toxicity control, temporal order and rescue or orthogonal replication. Avoid the phrase "proves the mechanism".
- Dose, exposure time, guide/construct, delivery, assay conditions and experimental model are experimental-design factors, not ordinary machine-learning hyperparameters. Never optimize them against a desired phenotype and then report the selected condition as confirmatory evidence.
- Distinguish biological replicates, technical replicates, fields, sections, wells, cells and repeated measurements. Inference follows the independently assigned biological or experimental unit.
- Be constructive. For every defect, give the criterion, minimum feasible repair, stronger option, resource consequence, closure evidence and the claim that survives without the repair.
- Label evidence as
PLANNED,AUTHOR_REPORTED,PARTLY_VERIFIED,VERIFIEDorNOT_ASSESSABLE. Never invent conditions, sample sizes, results, effect sizes or failures.
Choose the entry mode
| Mode | Use when | Required emphasis |
|---|---|---|
| plan | no experiment has been run | hypothesis discrimination, model/assay choice, controls, dose-time logic, unit, feasibility, decision and stop rules |
| audit | a protocol, Methods, notebook, images or result files exist | provenance, assignment, controls, independence, deviations, missing evidence, reproducibility and atomic repairs |
| mentor | a learner asks what to do next | explain why each design choice matters; give minimum, stronger and resource-limited routes without inventing a local SOP |
| interpret | real experimental outputs are supplied | target engagement before phenotype, magnitude/uncertainty, concordance, discordance, failure modes and claim ceiling |
| writing-handoff | the design and results are fixed | auditable Title/Abstract/Methods/Results/Figure/Supplement placement, negative evidence, claim ceiling, source-data map and separate data/code/reagent availability jobs |
Also declare system: tissue | cell-line | organoid | animal | mixed, intervention: none | genetic | pharmacological | environmental | mixed, and the affected Claim IDs.
Minimal intake
Record what is supplied and mark the rest AUTHOR_INPUT_NEEDED:
- biological question, proposed mechanism, affected imaging/molecular claim and at least one competing explanation;
- human cohort or discovery evidence motivating the experiment, including whether it is merely correlational;
- experimental system, biological source, authentication/status checks and relevance to the intended population;
- intervention, delivery, candidate dose/exposure levels and timing, with rationale source;
- primary target-engagement readout, primary phenotype, secondary readouts and measurement time;
- positive, negative, vehicle/vector, sham, batch and rescue controls that are actually available;
- independently assigned unit, planned biological replicates, technical structure, randomization, blinding, exclusions, attrition and stopping logic;
- local biosafety/ethics approvals, equipment, tissue, staff, budget, calendar and validated SOP constraints;
- supplied protocols, raw/processed outputs, laboratory records and exact artifact locations.
Required workflow
- Open the mechanism experiment workflow.
- Convert the claim into a causal contrast:
proposed mechanism versus competing explanation -> discriminating intervention/measurement -> expected patterns -> failure interpretation. - Select the weakest experimental system that can answer the question, then state its translation limit. Do not choose complexity for prestige.
- Build the assay and control matrix. Separate target engagement, phenotype, toxicity/viability, specificity and technical validity readouts.
- Declare the independent unit, allocation, blocking, randomization, blinding, replicate hierarchy, analysis population, exclusions, missingness/attrition and batch structure before outcomes.
- Order evidence in time: characterize model -> establish exposure/delivery -> demonstrate target engagement -> measure phenotype -> test specificity/rescue -> replicate or translate.
- Define dose and time as a scientific design problem. State rationale, permissible range and decision rule; never propose a universal concentration, duration or reagent recipe without a verified local/source protocol.
- Create minimum, stronger and resource-limited routes. Each route must state the question it can answer, extra cost/time/tissue and resulting claim ceiling.
- Apply the quality gates below. A fatal gate cannot be averaged away by attractive downstream results.
- Use the experiment plan, local assay capability profile, case-level mechanism-validation experience registry and failure/decision log when durable artifacts are useful. A local case may inspire a check, but never becomes a universal rule without external evidence.
Quality gates
| Gate | Pass question | If unresolved |
|---|---|---|
| E0 Claim | Is there a falsifiable claim, competing explanation and discriminating observation? | stop assay shopping; rewrite the question |
| E1 Model | Does the system contain the target/process and represent the intended biology? | limit to feasibility or choose another model |
| E2 Unit | Is the independently assigned unit explicit and protected from pseudo-replication? | redesign allocation and inference |
| E3 Intervention | Are identity, delivery/exposure, timing, specificity and assignment auditable? | no causal interpretation |
| E4 Engagement | Is target engagement measured independently of the headline phenotype? | phenotype cannot be attributed to the target |
| E5 Controls | Do controls distinguish baseline, delivery/vehicle, off-target, batch, assay failure, toxicity and rescue? | state the unresolved alternative explanation |
| E6 Analysis | Are endpoint, analysis population, exclusions, multiplicity and uncertainty prespecified at the correct unit? | exploratory only; hand off inference design |
| E7 Reproducibility | Are independent repeats, batch transfer, deviations, raw evidence and failure logs available? | do not claim robust replication |
| E8 Translation | Does cross-model or human evidence support the intended scope, with discordance explained rather than hidden? | constrain translation and clinical language |
Return PASS, CONDITIONAL or STOP per gate and per claim. STOP is appropriate for absent
target identity, unusable experimental unit, missing intervention contrast, outcome-selected
conditions, or an ethics/biosafety barrier.
Assay and system routing
| Route | Primary design job | Minimum discriminators | |---|---|---| | IHC / multiplex IF | localize and quantify cell/marker relationships in tissue | validated marker/assay controls, ROI sampling, scoring/blinding, section/block/patient hierarchy, batch control | | tissue molecular assay | test abundance, activation or pathway state in matched tissue | tissue provenance, purity/composition, pre-analytics, reference/normalization, batch and patient-level unit | | cell line | establish tractable perturbation-response logic | identity/contamination status, context relevance, multiple independent models where feasible, vehicle/vector and toxicity controls | | organoid / ex vivo | preserve patient or tissue context while enabling intervention | donor-level unit, derivation/selection bias, passage/culture state, matched control and cross-donor replication | | animal model | test system-level temporal or microenvironmental consequences | allocation unit, randomization/blinding, sex/age/model relevance, attrition, welfare endpoints and translation limit | | genetic perturbation | test target-specific necessity or sufficiency | independent guides/constructs, non-targeting control, editing/knockdown and protein/function engagement, off-target strategy and rescue | | pharmacological perturbation | test target/pathway modulation under exposure | vehicle, exposure and selectivity rationale, target engagement, toxicity, orthogonal agent or genetic triangulation | | rescue / reversal | distinguish target-specific effect from perturbation artifact | rescue construct/agent validity, expression/exposure comparability, timing, baseline effect and blinded analysis |
Scientific red lines
- Human imaging-omics association motivates experiments but is not biological validation by itself.
- Staining intensity, pathway score, ligand-receptor prediction or colocalization alone does not establish activity, directionality or causality.
- A single guide, inhibitor, cell line, donor, batch, section, animal or experiment cannot establish general mechanism. Repeated fields/wells/cells do not repair absent biological replication.
- Target engagement must precede attribution of phenotype. Toxicity, stress response, delivery artifact and batch must remain explicit competing explanations.
- Rescue is informative only when it addresses the same target/process and does not introduce a new uncontrolled exposure. A post hoc rescue chosen because it worked is exploratory.
- Do not convert exploratory dose/time screening into confirmatory evidence without a freeze and an independent evaluation.
- Do not prescribe a reagent concentration, animal dose, guide sequence, incubation time, sample size or stopping boundary as universally correct. Use source-backed and locally validated values.
- Do not provide clinical treatment advice. Human-subject, animal, genetic manipulation, hazardous material and biosafety decisions require the appropriate local review and trained personnel.
Output contract
Return the shortest complete package needed, in this order:
Scope and evidence boundary— mode, system/intervention, Claim IDs, supplied evidence and what remains planned or unverifiable.Mechanism contrast— proposed mechanism, competing explanations, causal graph in words, discriminating observations and current association/causal ceiling.Minimum / stronger / resource-limited routes— question answered, controls, extra tissue/time/cost and resulting claim ceiling.Experiment matrix— model, intervention, comparator, allocation unit, dose/time rationale, target-engagement readout, phenotype, toxicity/specificity readout and expected/falsifying pattern.Control and replicate matrix— positive/negative/vehicle/vector/sham/batch/rescue controls, biological versus technical hierarchy, randomization and blinding.Feasibility and governance— local capability status, tissue/reagent/equi
Truncated for display — read the full file on GitHub.
Related Skills
taste-skill
116.8kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
ai-job-search
72.1kThe job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.
claude-howto
47.3kA visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value.
Claude-Code-Game-Studios
39.7kTurn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
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.
