What this resource does
Core uses
About this resource
LabClaw is a modular collection of 240 OpenClaw-compatible Skills covering biology, lab automation, drug discovery, medicine, data science, literature research, and scientific visualisation.
This page groups representative academic components by task; review the repository for the complete inventory.
Literature search and review
Includes Skills for academic and biomedical search, citation management, literature review, patents, grants, and scientific writing.
- Typical inputs
- Research topic or question; Source criteria, date limits, and available database access
- Output
- Component-dependent search results, evidence notes, citations, or review text.
Best for: Finding and organising literature before a biomedical research task.
Skills for this task
pubmed-search
Searches PubMed for a defined biomedical research question.
literature-review
Organises a documented literature-review workflow.
Biomedical and laboratory analysis
Provides separate Skills for bioinformatics, single-cell analysis, laboratory automation, drug discovery, and clinical research tools.
- Typical inputs
- Selected domain Skill; Compatible data, protocol, software, or instrument access
- Output
- A component-dependent analysis, protocol, database result, or tool-specific artifact.
Best for: Using one named biomedical tool or workflow with suitable domain expertise.
Skills for this task
scanpy
Guides a standard single-cell RNA-seq analysis with Scanpy.
rdkit
Supports cheminformatics and molecular operations with RDKit.
Statistics and scientific figures
Contains Skills for statistical analysis, machine learning, and publication-oriented plotting.
- Typical inputs
- Dataset, variable definitions, analysis question, and figure requirements
- Output
- A component-dependent analysis result, code artifact, or scientific figure.
Best for: Analysing and presenting checked research data.
Skills for this task
statistical-analysis
Guides a general statistical-analysis workflow.
scientific-visualization
Creates publication-oriented scientific visualisations.
Use boundaries
Limits and checks
Component variation
Installation and risk cannot be assessed once for the whole collection.
Read the selected SKILL.md and its linked tool documentation before use.
Sensitive biomedical data
Configured agents or external services may receive sensitive material.
Use approved environments and follow institutional data-governance rules.
Method validity
Technically completed output may still be scientifically inappropriate.
Review assumptions, controls, methods, and interpretation with a domain specialist.
More boundaries
- Do not assume every LabClaw Skill supports every biomedical domain, file type, instrument, or database.
- A Skill cannot run a tool or access a dataset when the required package, account, API, hardware, or data is unavailable.
- Clinical, laboratory, statistical, and scientific conclusions still require qualified human review.
Technical details
- Resource type
- Skill Collection
- Source last updated
- 19 Mar 2026
- Last verified
- 15 Jul 2026
- Licence
- See source
- Access
- Publicly available
- Additional costs
- Platform terms or usage limits may apply. API usage fees may apply for selected components. External services, software, compute, or data access may have separate costs.
- Skill instruction language
- English
- Documentation language
- English
- Repository languages
- Markdown
- Dependencies
- OpenClaw; Component-specific research packages, APIs, accounts, instruments, or datasets
- Review status
- Quick test completed
Continue exploring
Related resources
Official package contents
Files
Read the official English instructions without leaving this profile. This view includes the English files represented by this profile. Non-English source files, supporting code, assets, tests, and references may not be shown.
Official package files
6 files available here
skills6
bio/scanpy1
general/statistical-analysis1
literature/literature-review1
literature/pubmed-search1
pharma/rdkit1
visualization/scientific-visualization1
Read on this page
Select a file
Choose a file from the tree to read its official instructions without leaving this page.
Independent testing
Evaluation report
See exactly what was tested, which checks were used, and what the result does not cover.
Evaluations are mainly based on automated AI-assisted checks. Results are for reference only.
Evaluation coverage
2 components tested separately
Each score applies only to the named component, cases, application and model. The scores are not combined into a Repository rating.
4 named components were not tested
pubmed-searchRequires a configured external API or service account. Not tested in this review. Contact us if you would like this capability tested.literature-reviewNot tested in this review.scanpyrdkit
Tested component 1statistical-analysis96 / 100 · One example
One example
Comparing two groups of measurements
The treatment group averaged 15.13 mg/L and the control group 12.50 mg/L. The reported difference was 2.63 mg/L. Recalculating the supplied data reproduced the reported numbers. Equal values overlap in the chart, and one additional distribution-check plot described in the guide was not produced.
Codex CLI · gpt-5.6-luna · 4 Sept 2026
What we asked it to do
Compare concentration measurements from two independent groups, calculate the difference and its uncertainty, and return a chart plus a short results paragraph.
Exact request used in the test
Analyze the two independent groups using a two-sided Welch t-test at alpha 0.05. Report group sizes, means, standard deviations, mean difference defined as treatment minus control, Welch t, degrees of freedom, two-sided p value, Hedges' g, and a 95% confidence interval for the mean difference. Check distribution and variance assumptions, create a box-and-point diagnostic chart, and write a 100-150 word APA-style result without causal language.
Original data
Sixteen synthetic concentration measurements, in mg/L. The groups were specified as independent.
- Control · 8 measurements
- 10, 12, 11, 13, 12, 14, 13, 15
- Treatment · 8 measurements
- 12, 14, 13, 16, 15, 17, 16, 18
Reported results
| Group | Measurements | Average | Standard deviation |
|---|---|---|---|
| Control | 8 | 12.50 mg/L | 1.60 mg/L |
| Treatment | 8 | 15.13 mg/L | 2.03 mg/L |
Difference in averages: 2.63 mg/L. The treatment average was higher. The reported 95% confidence interval for the difference was 0.65 to 4.60 mg/L.
This interval describes uncertainty in the estimated difference. The result does not show that the treatment caused the difference.
Statistical test values
- Welch t
- 2.869
- Degrees of freedom
- 13.285
- Two-sided p value
- 0.0129
- Hedges' g (effect size)
- 1.356
Read the original results paragraph · 143 words
The control group (n = 8) had a mean concentration of 12.50 mg/L (SD = 1.60), and the treatment group (n = 8) had a mean of 15.13 mg/L (SD = 2.03). A two-sided Welch independent-samples t-test indicated that the treatment-minus-control mean difference of 2.63 mg/L was statistically significant, t(13.28) = 2.87, p = .0129, Hedges’ g = 1.36. The 95% confidence interval for the mean difference was [0.65, 4.60] mg/L. Shapiro–Wilk tests indicated no evidence of non-normality in control (W = 0.983, p = .975) or treatment (W = 0.974, p = .925) values, and Levene’s test indicated no evidence of unequal variances (F(1, 14) = 0.548, p = .471). Welch’s procedure was retained as specified. These results describe an association between group membership and observed concentration in this independent-group sample; they do not establish causation or generalize beyond the measured groups.
What the result shows
The reported numbers match the data
Both groups contain eight measurements. Recalculation reproduced the group averages, spread, mean difference, test result, effect size and confidence interval. The interval for the difference is 0.65 to 4.60 mg/L; it does not include zero.
The paragraph meets the length request
The original response has 143 words, within the requested 100-150 range. It reports the uncertainty and explicitly avoids claiming that the treatment caused the difference.
Repeated values hide behind one another
The chart uses all 16 measurements, but only 13 distinct dot positions are visible. For example, the two control values of 12 sit on top of one another. Offsetting the dots would make the repetitions easier to see; the original chart below has not been altered.
The assumption check was not fully visual
The response reports numerical distribution and variance checks. It does not include the additional Q-Q plot described in the official guide, which helps a reader inspect the distribution. With only eight values per group, a non-significant normality test does not establish that the data follow a normal distribution.
How this example was scored
This score covers one selected example from six earlier test cases. The original input and output have not been changed. Only checks applicable to this example contribute to the score; the other five cases do not. This was a review of retained results, not a new Skill run.
Five scoring dimensions
Passed checks receive 1 point, partly met checks 0.5, and failed checks 0. Checks outside this example are excluded, not counted as passes. The five dimension percentages are combined using the weights shown below.
Did it complete the task?
95.0%The requested calculations and 143-word paragraph are present. The chart loses clarity where equal values overlap.
Did it run without problems?
87.5%The calculation and plot completed. A further distribution-check plot described in the guide was not produced.
Did it preserve and handle the supplied information responsibly?
100.0%Recalculation matched the numbers, all 16 values were retained, and the text did not claim that the treatment caused the difference.
Did it do the documented task in the tested setup?
100.0%The two-group task worked in this recorded Codex setup. This does not establish support for other models or applications.
Were the setup and use instructions clear enough to follow?
100.0%The instructions identify the inputs, Python packages and statistical reporting steps needed for this example.
Limits of this example
- These are synthetic test measurements, not findings from a real experiment.
- The test method was supplied in the request. This example does not show whether the Skill would choose an appropriate test by itself.
- This score does not cover missing-data handling, paired or repeated measurements, larger datasets, regression, or other AI models.
- PubMed search is a separate LabClaw component and was not run here; it requires configured external access.
Tested component 2scientific-visualization100 / 100 · One example
One example
Drawing a two-panel figure with uncertainty ranges
The component produced the requested two-panel figure. All 12 plotted means and their interval limits match the supplied table. The long labels and shared legend remain visible in both output formats.
Codex CLI · gpt-5.6-luna · 4 Sept 2026
What we asked it to do
Turn a supplied table into two side-by-side panels, showing two groups, their averages and uncertainty ranges. Keep the long names readable and export the figure.
Exact request used in the test
Create a horizontal two-panel interval figure. Panel A is 'Gene expression'; Panel B is 'Protein abundance'. Plot Control and Treatment means with asymmetric 95% confidence intervals, distinct Okabe-Ito colors and marker shapes, one shared legend, lowercase panel labels a and b, x-axis 'Normalized response (AU)', and the supplied long biomarker labels. Export a 7-by-4 inch PNG at 300 dpi and PDF.
Original data
A synthetic table with six biomarker names, two groups per name, and a supplied mean plus lower and upper 95% interval limits for each group. The values are for testing the drawing task, not findings from real research.
View the 12 supplied values and intervals
| Panel and biomarker | Group | Mean | Lower limit | Upper limit |
|---|---|---|---|---|
| A · Interleukin-6 receptor alpha | Control | 2.0 | 1.7 | 2.3 |
| A · Interleukin-6 receptor alpha | Treatment | 2.8 | 2.4 | 3.2 |
| A · Tumor necrosis factor receptor 1 | Control | 3.1 | 2.8 | 3.5 |
| A · Tumor necrosis factor receptor 1 | Treatment | 3.7 | 3.2 | 4.1 |
| A · Vascular endothelial growth factor A | Control | 1.5 | 1.2 | 1.9 |
| A · Vascular endothelial growth factor A | Treatment | 2.2 | 1.8 | 2.6 |
| B · Signal transducer and activator 3 | Control | 4.0 | 3.6 | 4.4 |
| B · Signal transducer and activator 3 | Treatment | 4.9 | 4.4 | 5.3 |
| B · Mitogen-activated protein kinase 1 | Control | 2.6 | 2.2 | 2.9 |
| B · Mitogen-activated protein kinase 1 | Treatment | 3.3 | 2.9 | 3.8 |
| B · Nuclear factor kappa B subunit 1 | Control | 3.5 | 3.0 | 3.9 |
| B · Nuclear factor kappa B subunit 1 | Treatment | 4.2 | 3.8 | 4.7 |
Generated figure
Circles and squares mark the supplied group means; horizontal lines show their supplied uncertainty ranges. The Skill drew these values rather than calculating them from raw observations. Original output, unchanged.
What the result shows
The numbers were plotted in the right places
All 12 means and 24 interval endpoints match the supplied table. The code subtracts each lower bound from its mean and each mean from its upper bound to draw the two sides of each interval.
The layout handles long names
Six long biomarker names fit without being cut off in the original image. Each panel has a title and letter, and a single legend identifies both groups. Different marker shapes distinguish the groups as well as their colours.
Both requested formats were produced
The PNG is 2100 by 1200 pixels. The PDF opens as one 7-by-4-inch page and has the same visible layout.
How this example was scored
This score covers one selected example from four earlier cases. It checks the requested figure and data preservation, not every chart type or scientific validity of the supplied intervals. One missing-data check is outside this example and is excluded. The original output is unchanged; this is not a new Skill run.
Five scoring dimensions
Passed checks receive 1 point, partly met checks 0.5, and failed checks 0. Checks outside this example are excluded, not counted as passes. The five dimension percentages are combined using the weights shown below.
Did it complete the task?
100.0%Both panels, all supplied values, the long labels and the requested exports are present.
Did it run without problems?
100.0%This local plotting run completed and produced files that opened.
Did it preserve and handle the supplied information responsibly?
100.0%The means and interval limits were preserved. They were supplied in the input, not estimated by the Skill.
Did it do the documented task in the tested setup?
100.0%The requested figure worked in this recorded Codex setup; other environments were not assessed.
Were the setup and use instructions clear enough to follow?
100.0%The guide provides plotting, layout and export instructions relevant to this example.
Limits of this example
- The task supplied the means and interval limits. Statistical estimation from raw data was not tested.
- Colours, markers, dimensions and panel layout were specified in the request; independent design choices were not assessed.
- A single successful example does not establish results for other datasets, chart types, models or journal requirements.
Want to see more tests?
Tell us which module or task you would like us to test next.