Skip to content
CAI
Software that uses CAICheck a score

K-Dense-AI/scientific-agent-skills

52.9

Adequate · 15 November 2026

150.2k

lines of production code

Python

primary language

5

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Features

Add pathway-enrichment skill for gene-set analysis

Introduces a new pathway-enrichment skill that performs over-representation analysis (ORA) and gene set enrichment analysis (GSEA) on gene lists. The skill provides a CLI helper script and comprehensive documentation covering method selection (ORA vs. GSEA), gene-ID mapping, background universe configuration, multiple-testing correction, and visualization using GSEApy 1.3.1, g:Profiler, and MSigDB.

skills/pathway-enrichment · high confidence

Add script to generate skill workflow diagrams via OpenRouter

A new Python script, generate\_skill\_image.py, has been added to the scripts directory to automatically create visual workflow diagrams for skills. The tool reads skill documentation (SKILL.md, references, etc.) and uses the OpenRouter API to first distill the content into a diagram description and then render it as an image using the openai/gpt-image-2.5-sunburst model. The generated images are saved to the docs/images directory and are designed to clearly explain a skill's steps, tools, and logic for users who have not previously interacted with the skill.

scripts · high confidence

Added installation inspection script for MarkItDown skill

A new \inspect\_installation.py\ script has been added to the MarkItDown skill to verify the environment without loading plugins or using the network. It checks the installed MarkItDown version against the target version (0.1.8), reports import health, lists declared extras and optional distributions (such as markitdown-ocr and azure-ai-documentintelligence), discovers registered plugins via entry points, and verifies the presence of required executables like ffmpeg and exiftool. The script supports both human-readable and JSON output and can exit successfully despite version mismatches if requested.

skills/markitdown/scripts · high confidence

Added version-checking script for IDC skill dependencies

A new \check\_version.py\ script has been added to the imaging-data-commons skill to verify that the \idc-index\ package meets the minimum required version (0.12.5) and to notify users when newer versions of the package or the skill itself are available. The script performs best-effort network checks for updates and provides specific installation commands if the current version is insufficient, ensuring users are aware of required updates without automatically modifying their environment.

skills/imaging-data-commons/scripts · high confidence

Exploratory Data Analysis skill introduces bounded, security-first local analysis with strict format support

The exploratory-data-analysis skill now provides a bounded, local-only workflow for inspecting authorized scientific files before modeling, enforcing a strict security boundary that rejects URLs, symlinks, pickle/joblib deserialization, and untrusted embedded instructions. It supports automated core inspection for CSV, TSV, and strict JSON, with optional bounded metadata inspection for NumPy (NPY/NPZ), HDF5, FASTA/FASTQ, and imaging formats (PNG/JPEG/TIFF) using pinned libraries (e.g., NumPy 2.5.3, h5py 3.16.0, Biopython 1.88, Pillow 12.3.0, tifffile 2026.9.20). The skill includes a detailed capability matrix, safe local I/O contracts (64 MiB default/512 MiB hard limits, atomic writes, tokenized identifiers), and rigorous EDA reasoning guidelines to prevent data leakage, automatic imputation/transformation, and causal overreach. Reference-only guidance is provided for complex formats like genomics (SAM/BAM/VCF), chemistry (PDB/SDF), and proteomics (mzML/mzIdentML), directing users to external validated tooling.

skills/exploratory-data-analysis · high confidence

HypoGeniC skill adds safety, reproducibility, and audit tooling

The HypoGeniC skill now includes a comprehensive safety and reproducibility framework for the \hypogenic\ 0.3.5 package. It introduces a local run policy (\run\_config.example.json\) that enforces explicit provider selection, cost/token caps, and mandatory user approval before any external LLM calls or model downloads. To ensure data integrity, the skill provides a dataset manifest schema and audit scripts that verify file hashes and detect cross-split data leakage. Additionally, it bundles local validation and inspection tools for configuration files, hypothesis banks, and evaluation results, ensuring that all operations are deterministic, bounded, and free from prompt injection or credential exposure risks.

skills/hypogenic · high confidence

Introduce Arbor skill for hypothesis tree state management

Adds the Arbor skill, providing a persistent hypothesis-tree state manager for Autonomous Optimization runs. The new \tree.py\ script handles the mechanical management of research state, including initializing runs, creating and linking nodes, recording evidence, pruning falsified branches, and rendering an 'Observe' projection for the coordinator. State is stored in \.arbor/tree.json\ and \.arbor/run.json\, ensuring auditable and consistent tracking of hypotheses, insights, and metadata during optimization cycles.

skills/arbor/scripts · high confidence

Introduce LiteParse skill for local document parsing

Added the \skills/liteparse\ skill, which provides local, layout-aware text extraction from PDFs, Office files, and images using the \liteparse\ library (version 2.15.0). This skill enables spatial text extraction with bounding boxes, OCR on scanned documents, and page screenshot generation for multimodal agents. The entry includes the skill definition (\SKILL.md\), detailed API and CLI references, and a Python batch-parsing script (\batch\_parse\_dir.py\) to support bulk document ingestion.

skills/liteparse · high confidence

Introduce Neuropixels analysis skill with SpikeInterface-based pipeline

Added a new Neuropixels analysis skill that provides a complete, SpikeInterface-driven workflow for processing extracellular recordings. The skill includes scripts for exploring raw data, preprocessing (bandpass filtering, bad channel removal, phase shift correction, and common median referencing), running spike sorters (Kilosort4, Kilosort3, SpykingCircus2, Mountainsort5), and computing quality metrics with configurable curation presets (Allen, IBL, strict). It also supports exporting results to Phy for manual curation and includes a comprehensive end-to-end pipeline script that integrates these steps.

skills/neuropixels-analysis/scripts · high confidence

Introduce OpenPIV skill for Particle Image Velocimetry analysis

Added the OpenPIV skill under skills/openpiv, providing a CLI runner and Python API to perform Particle Image Velocimetry (PIV) analysis on image pairs. The skill targets openpiv 0.26.1 and includes a bundled smoke-test script (run\_example.py) that validates the pipeline against OpenPIV's own test data. The runner supports dynamic intensity masking, configurable interrogation windows, vector validation, and outlier replacement, outputting velocity fields, flags, and visualizations. A post-processing helper (analyze.py) enables derived quantity computation, including vorticity, strain rate, and turbulence statistics, with correct handling of physical coordinate systems and grid spacing.

skills/openpiv · high confidence

Introduce PyDESeq2 skill for bulk RNA-seq differential expression

Added a new skill for performing bulk RNA-seq differential expression analysis using PyDESeq2 0.5.4. The skill includes a CLI driver script (\run\_deseq2\_analysis.py\) that handles count validation, formula design, explicit contrasts, Wald testing, and LFC shrinkage, along with comprehensive documentation covering statistical interpretation, normalization pitfalls, and API usage patterns.

skills/pydeseq2 · high confidence

Introduce autoskill skill for workflow-based skill discovery

The autoskill skill is now available to analyze a user's Screenpipe history windows, detect repeated research workflows, and propose new skill drafts or composition recipes for review. It requires a local Screenpipe daemon (default localhost:3030) and runs detection and embedding inference locally using sentence-transformers. The skill supports a local LLM via LM Studio by default, with optional opt-in cloud backends for Anthropic Claude and Microsoft Foundry. It includes a unified CLI with run, doctor, and promote subcommands, a configuration file for backend and clustering settings, and a privacy posture that redacts sensitive data before sending summaries to the LLM.

skills/autoskill · high confidence

Introduce bulk RNA-seq skill with reproducible quantification and DE handoff

Adds a new bulk RNA-seq skill that guides users from raw FASTQ files to differentially expressed genes. The skill supports two upstream paths: running the nf-core/rnaseq pipeline (Path A) or using standalone tools like STAR, Salmon, and featureCounts (Path B). It includes a Python bridge script (build\_counts\_matrix.py) to aggregate quantification output into a gene-level counts matrix compatible with PyDESeq2, along with samplesheet validation and detailed reference documentation on experimental design, QC, and statistical handoffs.

skills/bulk-rnaseq · high confidence

Introduce fail-closed validation scripts for clinical report artifacts

The \skills/clinical-reports/scripts\ directory now includes a suite of Python tools that enforce strict, fail-closed validation on structured clinical report manifests. These scripts—covering de-identification checklists, provenance traceability, terminology coding, consistency checks, and coverage for CARE, ICH E3, CONSORT, and SPIRIT guidelines—verify schema integrity, data classification, and safety gates without accessing or modifying the underlying clinical data. By requiring specific draft statuses (e.g., \BLOCKED\_INCOMPLETE\) and rejecting unverified or malformed inputs, the new scripts ensure that reports remain in a blocked, non-submission state until all structural and privacy requirements are met.

skills/clinical-reports/scripts · high confidence

Introduce ontology-term-resolution skill for validating and mapping scientific identifiers

A new skill has been added to resolve free-text scientific labels to ontology term IDs and validate existing CURIEs against the EBI Ontology Lookup Service (OLS4). The skill includes Python scripts (resolve\_terms.py, validate\_terms.py, lookup\_prefix.py, map\_terms.py) that use only the standard library to query OLS4, Bioregistry, Identifiers.org, ZOOMA, and Ontobee. It provides a decision procedure for mapping lab shorthand, checking prefix validity, and auditing metadata tables for obsolete or mismatched terms, ensuring that ontology identifiers are verified before being written to submission templates or metadata files.

skills/ontology-term-resolution · high confidence

New 13C metabolic flux analysis skill for steady-state isotope tracing

A new skill has been added to estimate intracellular metabolic fluxes from steady-state carbon-13 isotope-tracing measurements. This tool uses the \mfapy\ library (pinned to commit a10433af) for forward simulation and SciPy for constrained multistart fitting, allowing users to determine feasible flux estimates and assess which pathway fluxes are constrained by their labeling data. The skill includes a CLI (\scripts/mfa.py\) with commands for input validation, forward simulation, and flux fitting with profile diagnostics. It is designed for metabolic and isotopic steady-state conditions, supporting positional isotopomers and parallel tracer experiments, and comes with documented input contracts and synthetic worked examples (a two-route branch model and a TCA cycle demonstration) to guide usage.

skills/13c-metabolic-flux · high confidence

New AI-powered schematic generation and PDF quality checks for scientific posters and schematics

Added new scripts to the latex-posters and scientific-schematics skills to automate the creation and validation of scientific diagrams. The generate\_schematic.py scripts now use the Nano Banana 2 model with iterative refinement and quality reviews by Gemini 3.7 Flash, allowing users to generate diagrams for various document types (e.g., journals, posters) with configurable quality thresholds. Additionally, a new review\_poster.sh script provides automated preflight checks for PDF posters, verifying page dimensions, font embedding, and image resolution using Poppler tools, while example\_usage.sh demonstrates the new generation capabilities.

skills/latex-posters/scripts, skills/scientific-schematics/scripts · high confidence

New AlphaGenome skill for variant effect prediction and interpretation

Added the AlphaGenome skill, providing access to DeepMind's AlphaGenome model and the AlphaGenome Atlas for predicting and interpreting the molecular effects of single-nucleotide variants. The skill includes Python scripts (\atlas\_query.py\, \score\_variants.py\, \atlas\_link.py\) and comprehensive documentation (\SKILL.md\, \references/\) to query precomputed Atlas scores (AVI, track-level effects) for GRCh38 SNVs or run on-demand model predictions for variants, indels, and in silico mutagenesis scans. It requires the \alphagenome\ Python package (v0.9.0+) and an API key, enabling users to rank variants, resolve biological mechanisms via specific tissue tracks, and generate deep links to the Atlas portal.

skills/alphagenome · high confidence

New BioServices skill with validated API contracts and bundled analysis scripts

The bioservices skill has been added, targeting bioservices 1.16.0 and Python 3.9+. It includes comprehensive reference documentation detailing the correct API contracts for UniProt, KEGG, QuickGO, UniChem, and EMBL-EBI BLAST, correcting previous assumptions about deprecated methods and response formats. The skill also bundles scripts for batch identifier conversion, compound cross-referencing, and KEGG pathway analysis, providing users with validated workflows for protein annotation, chemical mapping, and pathway network extraction.

skills/bioservices · high confidence

New DeepChem molecular property prediction skill

Added a new DeepChem skill (version 2.0) that provides validated workflows for molecular property prediction, including SMILES featurization, scaffold holdouts, and explicit pretrained encoder transfer. The skill targets DeepChem 2.8.0 with Python 3.11 and includes bundled scripts for solubility regression, graph neural network training (GCN, GAT, AttentiveFP, MPNN, DMPNN), and transfer learning with Hugging Face models (ChemBERTa, MoLFormer) and GROVER. It also supplies detailed API reference documentation and review evidence to ensure scientific validity and correct usage of DeepChem's data loaders, splitters, and model contracts.

skills/pymc · high confidence

New DiffDock skill for protein-small-molecule docking

Added a new skill for DiffDock (v1.1.3) that predicts protein-small-molecule binding poses. The skill includes a comprehensive guide (SKILL.md) detailing setup, single-pair and batch inference workflows, and result analysis. It provides helper scripts to validate batch CSV inputs (checking for safe names, valid SMILES, and file existence) and to analyze results (extracting confidence scores and ranking poses). A custom inference configuration file is included to allow users to adjust sampling parameters, with clear documentation on how YAML settings override command-line arguments. The skill also documents the scientific scope, confidence interpretation, and validation requirements, emphasizing that confidence scores are for pose triage and not binding affinity.

skills/diffdock · high confidence

New ETE Toolkit and FlowIO skills for phylogenetic and cytometry analysis

Added new skill definitions and supporting assets for the ETE Toolkit (targeting ETE 4.4.0) and FlowIO (targeting 1.4.0). The ETE skill provides a comprehensive guide for analyzing, manipulating, and visualizing phylogenetic trees, including a migration path from ETE 3, API references, and validated Python scripts for tree operations and visualization. The FlowIO skill covers reading, inspecting, and writing Flow Cytometry Standard (FCS) files, with guidance on metadata handling and event extraction.

(repo-wide) · high confidence

New FlowKit cytometry analysis skill with CLI and documentation

Added a new FlowKit skill (version 1.1) for analyzing flow cytometry data, including a Python CLI script (\analyze\_gates.py\) that applies GatingML 2.0 strategies or FlowJo 10 workspaces to FCS files and exports gate counts, percentages, and provenance. The skill requires Python 3.13 and FlowKit 1.3.2, and includes reference documentation on compensation, transforms, gating hierarchies, and workspace result interpretation.

skills/flowkit · high confidence

New GeoPandas skill with local audit tools and safety guidelines

This change introduces a new GeoPandas skill (version 1.4) that provides guidance and local audit tools for Python workflows using GeoPandas 1.2.0. The skill includes a set of bundled CLI scripts (e.g., \vector\_inventory.py\, \crs\_reprojection\_plan.py\, \geometry\_validity\_report.py\) and reference documentation covering CRS management, data structures, geometric operations, spatial analysis, and visualization. It enforces a local-only intake policy, rejecting URLs, archives, and remote URIs, and requires explicit approval for any network access. The skill also details safety and privacy contracts, including handling sensitive coordinates, validating geometry validity, and ensuring correct CRS usage for planar measurements.

skills/geopandas · high confidence

New ISO standards readiness skill replaces ISO 13485 certification skill

The \iso-13485-certification\ skill has been replaced by a new \iso-standards-readiness\ skill (version 1.3) that broadens scope to include ISO 14971, ISO/IEC 17025, and ISO 15189. This update introduces a structured evidence-preparation workflow with new JSON templates for CAPA records, document registers, evidence manifests, and laboratory scope intakes, alongside Markdown templates for CAPA and document control procedures. It also adds a specific checklist for the FDA QMSR transition (effective 2026-02-02) and enforces strict boundaries to ensure the tool supports only draft evidence preparation for human review rather than making compliance or certification decisions.

skills/iso-standards-readiness · high confidence

New MATLAB and GNU Octave skill with R2026a-focused safety and reproducibility guidance

A new MATLAB skill (version 1.4) has been added to provide structured guidance for designing, reviewing, and migrating numerical workflows in MATLAB R2026a and GNU Octave 11.3.0. The skill introduces strict safety boundaries that prohibit the execution of untrusted code, MAT files, or MEX binaries without explicit approval, and it enforces static inventorying of dependencies and licenses before any runtime interaction. It includes detailed references on data import/export, graphics export (emphasizing \exportgraphics\), numerical reproducibility (tolerances, RNG state), and array indexing, alongside JSON templates for project and reproducibility manifests to standardize environment tracking. The skill also clarifies the partial compatibility between MATLAB and Octave, warning against assuming API or behavioral equivalence between the two runtimes.

skills/matlab · high confidence

New OMERO integration scripts for configuration validation, inventory, metadata export, and transfer planning

The OMERO integration skill now includes four new command-line scripts to support safer and more structured interactions with OMERO servers. The \validate\_config\ script allows users to verify endpoint and authentication settings locally without contacting the server, including optional DNS resolution. The \inventory\ script generates a bounded, read-only JSON inventory of OMERO objects (such as Images, Datasets, and Files) with configurable limits and optional name inclusion. The \export\_image\_metadata\ script exports bounded annotations and ROI geometry for specific image IDs, with options to include owner names, file names, and ROI labels. Finally, the \plan\_transfer\ script helps users create local import or export plans by scanning local paths or proposing OMERO export commands for specific images, enforcing strict bounds on file counts and directory depth to prevent excessive resource usage.

skills/omero-integration/scripts · high confidence

New OneKGPd skill for individual-level 1000 Genomes Project queries

Adds the \onekgpd\ skill, enabling queries against the 1000 Genomes Project dataset (3,202 whole-genome-sequenced individuals, GRCh38). The skill provides two offline and online components: \scripts/onekgpd\_api.py\ connects to the public \dnaerys\ service (v0.2.1) over TLS to select/count variants and samples, check homozygous-reference status, and calculate kinship; \scripts/onekgpd\_meta.py\ runs locally against a bundled \kgpe.json\ asset to query sample and population metadata. It requires Python 3.11+ and \uv\ to run.

skills/onekgpd · high confidence

New PK/PD modeling skill for pharmacokinetic and pharmacodynamic analysis

A new \pkpd-modeling\ skill has been added, providing a suite of Python-based tools for non-compartmental analysis (NCA), individual and population PK fitting, regimen simulation, exposure-response modeling, bioequivalence calculations, and scaling/first-in-human orientation. The skill includes CLI scripts for NCA, compartmental fitting, population PK dataset validation, regimen simulation, and exposure-response analysis, along with comprehensive documentation covering regulatory conventions (ICH M12, M13A/B), dataset standards, and software ecosystem guidance. It requires Python 3.12+ with NumPy 2+ and SciPy, and is designed for exploratory analysis without network access or proprietary software dependencies.

skills/pkpd-modeling · high confidence

New Pathogen Variant Surveillance skill for genomic data analysis

A new skill has been added to query public GenSpectrum LAPIS data for pathogen genomic surveillance, including lineage nomenclature, weekly sequence proportions, reporting delays, and mutation frequencies. The skill provides Python 3.11+ scripts (reporting\_lag.py, lineage\_prevalence.py, resolve\_lineage.py, mutation\_profile.py) that interact with 15 registered LAPIS instances across cov-spectrum.org, genspectrum.org, and pathoplexus.org. It supports SARS-CoV-2, influenza (H5N1, H3N2, H1N1pdm), RSV, mpox, measles, dengue, and several other pathogens, distinguishing sequence prevalence from infection prevalence and providing Wilson intervals for sampling uncertainty. The skill includes comprehensive documentation on data provenance, reporting delays, and interpretation caveats.

skills/pathogen-variant-surveillance · high confidence

New PrimeKG and PyTorch Lightning agent skills

Added two new agent skills: a PrimeKG skill that provides a local pandas-based query interface for the Precision Medicine Knowledge Graph CSV (including node search, neighbor retrieval, disease context, and path finding) with strict provenance and validation guidance, and a PyTorch Lightning skill that documents Lightning 2.6.6 usage with executable templates and reference guides covering LightningModules, Trainers, DataModules, callbacks, logging, distributed training, and best practices.

skills/primekg, skills/pytorch-lightning · high confidence

New PufferLib skill with version-aware guidance and safe defaults

Added a new PufferLib skill that provides version-aware guidance for reinforcement-learning environments, vectorization, policies, and training. The skill documents three distinct profiles: the native 5.0 source build, the published 3.0.0 PyPI release, and a pinned historical 4.0 source snapshot. It includes reference documentation for environment contracts, integration patterns, policy architectures, training workflows, and vectorization, along with bundled, dependency-free Python scripts for contract validation, vectorization benchmarking, and plan validation. The skill enforces safe defaults by restricting tool usage to allowlisted built-ins, requiring source review and hash verification for installations, and prohibiting the execution of unreviewed native extensions, ROMs, or checkpoints.

skills/pufferlib · high confidence

New PyOpenMS 3.6.0 skill with updated API bindings and CLI workflows

This change introduces a new \pyopenms\ skill that provides ready-to-run Python scripts and reference documentation for mass spectrometry workflows (proteomics and metabolomics) using pyOpenMS 3.6.0. The skill includes CLI tools for inspecting and converting MS data, feature detection and quantification, adduct annotation, identification post-processing, and chemistry calculations. It specifically addresses API changes in pyOpenMS 3.6.0, such as the move to nanobind bindings, updated method signatures (e.g., \Param.keys()\ returning strings, \IdXMLFile.load\ returning tuples), and corrected parameter handling (e.g., \--mz-tol-da\ for absolute m/z tolerance instead of ppm conversion). The skill ships with scripts like \align\_link\_quantify.py\ for multi-sample quantification and \accurate\_mass\_search.py\ for metabolite annotation using local HMDB databases, along with detailed reference guides on data structures, file I/O, and signal processing.

skills/pyopenms · high confidence

New PyTDC skill with verified API snapshot and safety controls

Adds a new PyTDC (Therapeutics Data Commons) skill that pins PyTDC 1.1.15 and CPython 3.11, and explicitly pins setuptools 80.9.0 to avoid runtime failures from the removed pkg\_resources module. The skill bundles CLI scripts for metadata discovery, dataset loading and splitting, benchmark evaluation, and molecular oracle scoring, all gated behind explicit user approval to control network downloads and disk usage. It includes reference documentation covering dataset registries, oracle side effects, split behaviors, evaluator metrics, and source verification, and enforces workspace-safe paths and bounded inputs to prevent unintended data exfiltration or storage bloat.

skills/pytdc · high confidence

New RDKit molecular analysis scripts for properties, similarity, and substructure filtering

Added four new Python scripts in the RDKit skills module: \_common.py provides shared utilities for reading molecular files (SMILES, SDF, MOL) while preserving source record indices; molecular\_properties.py calculates comprehensive descriptors including MW, LogP, TPSA, QED, and Lipinski/Lead-likeness rules; similarity\_search.py enables fingerprint-based screening (Morgan, RDKit, MACCS, etc.) with configurable thresholds and metrics; and substructure\_filter.py allows filtering molecules by SMARTS/SMILES patterns with built-in libraries for functional groups, rings, PAINS, and privileged structures.

skills/rdkit/scripts · high confidence

New RELSA severity assessment skill for animal welfare scoring and forecasting

Added a new skill that implements the RELSA (RELative Severity Assessment) methodology in Python to compute multivariate severity scores for laboratory animal studies and perform exploratory ARIMA-based score forecasting. The skill enables combining metrics like body weight, temperature, and clinical scores into a single severity index, supports defining attention/danger zones via kernel density estimation, and provides CLI and Python APIs for reference-relative scoring and time-point prediction, targeting compliance with 3Rs refinement and EU Directive 2010/63/EU reporting contexts.

skills/relsa-severity-assessment · high confidence

New Stable Baselines3 skill targeting version 2.9.0

This change introduces a new skill for Stable Baselines3 (SB3) version 2.9.0, providing comprehensive documentation and executable templates for training reinforcement learning agents. The skill covers core capabilities including algorithm selection (PPO, SAC, DQN, etc.), custom Gymnasium environment creation, vectorized environment usage, and callback-based monitoring. It includes a custom environment template, a training script with checkpointing and normalization support, and an evaluation script for reproducible model assessment. The skill requires Python 3.10+, PyTorch \>= 2.8, and stable-baselines3 2.9.0.

skills/stable-baselines3 · high confidence

New TimesFM forecasting examples with corrected quantile mapping and schema validation

The timesfm-forecasting skill now includes new illustrative examples for global-temperature and covariates-based retail forecasting, alongside a refactored core script (forecast\_csv.py). These examples use the TimesFM 2.5 model (package 3.0.2) and enforce a new output schema (2.5-deciles-v1) that correctly maps quantile indices (mean at 0, q10–q90 at 1–9). The scripts now validate forecast outputs to ensure quantiles do not cross and that the point forecast matches the median, rejecting legacy unvalidated JSON. The global-temperature example also provides new visualization tools, including an animated GIF and an interactive HTML dashboard, to show forecast evolution as more historical data is added.

skills/timesfm-forecasting/examples · high confidence

New Waypoint Bio skill for microbiome foundation models

Added the \waypoint-bio\ skill, providing documentation and scripts for Outpost Bio's Waypoint microbiome foundation models (Waypoint-6m, 45m, 170m), the Atlas pretraining corpus, and the Compass benchmark. The skill includes a CLI reference for embedding, fine-tuning, and benchmarking, along with Python API examples and utility scripts to convert MetaPhlAn, Kraken2, QIIME 2, and MGnify profiler outputs into the required waypoint format. It also documents known upstream limitations, such as tokenizer save issues and specific data formatting requirements.

skills/waypoint-bio · high confidence

New analytical method validation skill for regulatory compliance

Added a new skill that plans, executes, and documents analytical method validation and verification under ICH Q2(R2)/Q14, USP \<1220\>/\<1225\>/\<1226\>, ICH M10, CLSI EP, and ISO/IEC 17025. It provides Python scripts (using only the standard library) to generate framework-specific validation protocols, evaluate calibration response, accuracy, precision, detection/quantitation limits, and bioanalytical run acceptance, and compare methods. The skill also includes reference guides on framework selection, ICH M10 numeric criteria, ICH Q2(R2) structure, and compendial/CLSI/ISO sources, along with validation protocol and report templates to ensure pre-stated acceptance criteria and proper documentation.

skills/analytical-method-validation · high confidence

New citation and poster assets added

Added a BibTeX template file with illustrative examples for various publication types (articles, books, conference papers, etc.) and a citation quality checklist to help users validate their references. Additionally, added LaTeX poster templates using the baposter, beamerposter, and tikzposter packages, along with a poster quality checklist to guide users through compilation and visual inspection.

skills/citation-management/assets, skills/latex-posters/assets · high confidence

New clinical report scaffolds and safety checklists

The clinical-reports skill now includes a comprehensive set of JSON scaffolds and checklists to structure report creation and enforce safety/compliance workflows. These assets provide structured templates for case reports (CARE), clinical trial summaries (ICH E3), trial results (CONSORT), and safety aggregates, alongside specific scaffolds for laboratory, pathology, and radiology reports. Additionally, new manifests and checklists are provided for provenance tracking, terminology coding, consistency verification, de-identification (HIPAA Safe Harbor/Expert Determination), quality review, and trial protocol reporting (SPIRIT). Each template includes explicit safety notices, draft status flags, and review workflows to ensure that reports are populated only from verified sources and undergo appropriate clinical, privacy, and regulatory checks before release.

skills/clinical-reports/assets · high confidence

New evidence-first market research skill with auditable reporting templates

The \skills/market-research-reports\ area now provides a complete, evidence-first workflow for creating auditable market research reports. This includes a detailed operational guide (\SKILL.md\) enforcing strict claim-source mapping, uncertainty preservation, and ethical data collection. The skill ships with a suite of structured templates for managing evidence: CSV ledgers for sources and claims, JSON schemas for market sizing scenarios and forecast sensitivity, and a LaTeX package (\market\_research.sty\) with specific styling for evidence, calculations, assumptions, and limitations. Reference documents define the evidence model, citation integrity rules, and analysis patterns for TAM/SAM/SOM reconciliation, ensuring that all generated reports are traceable and reproducible.

skills/market-research-reports · high confidence

New experimental-design skill with DOE and randomization scripts

A new \experimental-design\ skill has been added to help users plan studies before data collection, covering design selection, randomization, blocking, and factorial/DOE layouts. The skill includes Python scripts (\scripts/doe\_designs.py\ and \scripts/randomization.py\) that wrap \pydoe==1.5.0\ to generate reproducible allocation schedules and design matrices in real units, along with comprehensive reference documentation (\references/\) explaining design types, aliasing, sequential/adaptive methods, and pseudoreplication pitfalls. It requires Python \>=3.12 with \numpy\, \pandas\, and \pydoe\.

skills/experimental-design · high confidence

New genomic-coordinates skill for format conversion and validation

Added the genomic-coordinates skill, which provides documentation and Python scripts to convert genomic intervals between coordinate conventions (e.g., BED 0-based half-open to GFF/VCF 1-based inclusive), normalize variant representations (left-aligning and trimming indels), and audit files for assembly or naming mismatches. The skill includes scripts for converting coordinates, normalizing variants against a reference FASTA, checking contig naming and assembly compatibility (distinguishing GRCh37, hg19, GRCh38, and T2T), and auditing interval files for convention violations, all using only the Python standard library.

skills/genomic-coordinates · high confidence

New gget analysis scripts for batch, enrichment, and gene workflows

Added three new executable scripts to the gget skills directory: \batch\_sequence\_analysis.py\ for running BLAST, alignment, and optional structure-prediction placeholders on multiple FASTA sequences; \enrichment\_pipeline.py\ for performing comprehensive gene-set enrichment analysis across multiple databases (e.g., KEGG, GO) with summary reporting and expression data retrieval; and \gene\_analysis.py\ for single-gene analysis including search, annotation, sequence retrieval (with FASTA output normalization), tissue expression, correlation, and disease/drug associations via Open Targets. These scripts provide structured, command-line accessible workflows for common bioinformatics tasks using the gget library.

skills/gget/scripts · high confidence

New hypothesis-generation skill with structured templates and safety boundaries

A new hypothesis-generation skill has been added to the system, providing a comprehensive framework for formulating evidence-bounded scientific questions and testable research plans. The skill includes a detailed SKILL.md guide that enforces strict safety and ethical boundaries, requiring local processing of sensitive data and prohibiting the presentation of hypotheses as established facts. It introduces a structured workflow covering scope definition, observation freezing, research question framing, and the generation of rival hypotheses with discriminating predictions. To support this process, the skill bundles several new asset templates: JSON schemas for hypothesis records, falsification controls, operationalization checks, and search boundaries, along with CSV templates for evidence and prediction matrices. A preregistration scaffold is also provided to help users document their analysis plans transparently. The skill is designed to be deterministic and local-only, relying on Python 3.11+ standard libraries and requiring no external models or network access.

skills/hypothesis-generation · high confidence

New infographic generation script with iterative quality review

A new script, generate\_infographic.py, has been added to the infographics skill. It enables users to generate professional infographics using the Nano Banana 2 model (Gemini 3.1 Flash Image) with smart iterative refinement. The script includes a quality review step using Gemini 3.7 Flash, regenerating content only if the quality falls below a threshold determined by the document type (e.g., marketing, report, social). It supports 10 infographic types (such as statistical, timeline, process) and 8 style presets (such as corporate, healthcare, technology), along with colorblind-safe palettes. The script handles API key resolution from environment variables or .env files and forwards a minimal set of environment variables to the subprocess for security.

skills/infographics/scripts · high confidence

New lab hardware CAD skill for parametric design and standard-compliant fabrication

Added a new 'lab-hardware-cad' skill that enables the design of custom laboratory hardware (such as microfluidic chips, optomechanical mounts, and animal-behavior rigs) as parametric Python models using build123d 0.13.0. The skill enforces a rigorous workflow where designs must interface with published standards (e.g., ANSI/SLAS microplate dimensions, optical breadboard grids) defined in a new \assets/standards.json\ database. It includes reference documentation for fabrication limits, material compatibility, and regulatory considerations, along with scripts to automatically verify geometric fit, clearance, and standard compliance before fabrication artifacts (STEP, STL, DXF) are exported.

skills/lab-hardware-cad · high confidence

New literature-review scripts for PDF generation, schematic creation, and citation verification

The literature-review skill now includes four new Python scripts to extend its capabilities. \generate\_pdf.py\ allows users to render reviewed Markdown documents into PDFs using Pandoc and XeLaTeX, with support for local CSL citation styles, bibliographies, and templates. \generate\_schematic.py\ enables the creation of scientific diagrams via the Nano Banana 2 AI model, featuring document-type-aware quality thresholds and iterative refinement. \search\_databases.py\ provides local processing of normalized search records, including deduplication, ranking, and formatting into JSON, Markdown, or BibTeX. Finally, \verify\_citations.py\ checks DOI registration status and retrieves metadata from Crossref to assist in manual citation review.

skills/literature-review/scripts · high confidence

New marine carbonate chemistry skill for seawater speciation calculations

Introduces a new skill that solves seawater carbonate chemistry using PyCO2SYS 1.8.3.4. The skill provides a Python script (\scripts/solve\_carbonate.py\) that accepts paired measurements (such as total alkalinity and dissolved inorganic carbon) from a CSV file and outputs a speciation table, mineral saturation estimates, and uncertainty propagation. It supports converting measurements between laboratory and in-situ conditions, handles pH scale distinctions, and generates a provenance manifest to ensure reproducibility.

skills/marine-carbonate-chemistry · high confidence

New pacsomatic skill script for pipeline validation and execution

A new script, run\_pacsomatic.py, has been added to the skills/pacsomatic/scripts directory. This tool allows users to prepare, validate, and launch the nf-core/pacsomatic pipeline. It includes robust environment detection for Java (versions 17-26) and Nextflow (minimum version 24.4.2), supports multiple HPC schedulers (LSF, Slurm, PBS, SGE), and handles local or cloned pipeline repositories with version control integration.

skills/pacsomatic/scripts · high confidence

New peer-review skill with structured intake, reporting-guideline selection, and local-only processing

The peer-review skill (version 2.4) introduces a rigorous, evidence-bounded workflow for preparing constructive review drafts and structured manuscript assessments. It enforces a mandatory safety boundary requiring explicit authorization, conflict disclosure, and venue-policy checks before any analysis, with a strict default to local-only processing that prohibits sending unpublished content to external services. The skill provides a structured intake gate (validated by a local CLI) and a multi-step review workflow covering scope establishment, reporting-guideline selection (using a new catalog of guidelines like CONSORT 2025, SPIRIT 2025, PRISMA 2020, and various AI-specific extensions), claim-evidence mapping, and statistical/reproducibility auditing. It includes comprehensive templates for intake, study profiles, claim-evidence matrices, reporting checklists, and review scaffolds, along with reference documents on common methodological issues and ethical review practices to support accountable human review.

skills/peer-review · high confidence

New phylogenetics and scVelo analysis skills with validated workflows

Added two new skills: a phylogenetics skill that automates MAFFT alignment, IQ-TREE 3/FastTree inference, and ETE3 visualization (version 1.5), and an scVelo skill for RNA velocity analysis using scVelo 0.3.4 on Python 3.13 (version 2.0). Both skills include detailed SKILL.md documentation, reference guides for their respective tools (IQ-TREE inference, velocity models), and Python scripts (phylogenetic\_analysis.py, rna\_velocity\_workflow.py) that enforce strict input validation, handle dependency checks (e.g., NumPy 2 compatibility for scVelo), and produce structured outputs with diagnostic reporting.

skills/phylogenetics · high confidence

New primer-design skill for PCR and RT-qPCR assay planning

A new 'primer-design' skill has been added to the system, providing tools to design and audit PCR and RT-qPCR primers using the Primer3 engine (via primer3-py 2.3.1). This skill enables users to generate candidate primer pairs, assess their thermodynamics (Tm, hairpins, dimers), and screen for specificity against local or BLAST-based references. The package includes a structured workflow for defining assay contracts, handling specialized cases like exon-junction or multiplex assays, and generating traceable evidence reports. It requires Python 3.11+ and local installation of NCBI BLAST+ for advanced screening capabilities.

skills/primer-design · high confidence

New pydicom skill with bounded local processing and privacy workflows

Added a new pydicom skill (version 1.4) that provides local-only scripts for reading, inspecting, and transforming DICOM datasets, including metadata extraction, technical inventory, pixel rendering, and pseudonymization. The skill pins pydicom 3.0.2 (which addresses [CVE redacted]) and includes helper CLIs with strict size/element limits, explicit privacy boundaries, and de-identification audit capabilities.

skills/pydicom · high confidence

New pymatgen skill with pinned scientific packages and safety-focused CLI tools

This change introduces a new pymatgen skill that pins the verified snapshot to pymatgen 2026.9.24, pymatgen-core 2026.9.23, and mp-api 0.46.5. It provides comprehensive reference documentation for core objects, symmetry, phase diagrams, I/O, and the Materials Project API, alongside bundled CLI scripts (composition/structure validator, artifact manifest generator, symmetry sensitivity report, and MP query tool) that enforce strict input validation, byte limits, and provenance tracking to mitigate risks such as untrusted CIF parsing and accidental credential exposure.

skills/pymatgen · high confidence

New pymoo optimization skill added

A new skill for the pymoo multi-objective optimization library has been added, providing documentation and reference materials for version 0.6.2. The skill covers core concepts, algorithm selection (including NSGA-II, NSGA-III, and MOEA/D), constraint handling, and decision-making workflows. It includes detailed reference guides for genetic operators, parallelization strategies, benchmark problems, and visualization techniques, along with executable scripts for custom problem definitions and quick-start workflows.

skills/scanpy · high confidence

New resource detection and workload planning skill

A new 'get-available-resources' skill has been added to provide conservative, privacy-preserving resource detection and workload planning. The skill includes a Python-based detector that gathers a JSON snapshot of host inventory, process affinity, cgroup/container limits, scheduler allocations, and accelerator visibility without running stress tests or exposing sensitive host details. It also provides a workload planner that uses this snapshot to calculate safe worker and memory ceilings, and an accelerator diagnostic tool that generates read-only probe plans for CUDA, ROCm, and Metal backends. The skill is documented with a detailed schema, semantic rules for interpreting resources, and verified source references.

skills/get-available-resources · high confidence

New scientific agent skills and safety infrastructure

This update introduces several new capabilities for scientific research, including AlphaGenome for genomic variant analysis, Exa Search for web and publication discovery, genomic-coordinates for converting between genomic interval formats, and analytical-method-validation for statistical quality checks. It also adds new safety and validation scripts for clinical decision support, clinical reports, and DNAnexus integrations to enforce stricter input bounds, prevent PII leakage, and validate application manifests.

scientific-agent-skills · high confidence

New scientific agent skills for gene network inference and medicinal chemistry filtering

Added new skill definitions and execution scripts for Arboreto (gene regulatory network inference using GRNBoost2/GENIE3 with pinned Python 3.11 and Dask compatibility) and Medchem (molecular filtering using drug-likeness rules, structural alerts, and complexity metrics with pinned medchem 2.1.1, datamol 0.13.0, and RDKit 2026.3.6). Each skill includes a SKILL.md with installation, usage, and troubleshooting guidance, reference documentation for algorithms and APIs, and bundled wrapper scripts (basic\_grn\_inference.py, filter\_molecules.py) that validate inputs and run the respective libraries.

skills/arboreto, skills/medchem, skills/pyhealth · high confidence

New scientific analysis skills for Cantera, CellProfiler, and MAGeCK

Added new reusable analysis skills for Cantera (homogeneous chemical reactor ignition delay calculations), CellProfiler (microscopy pipeline execution and nuclei measurement QC), and MAGeCK (CRISPR pooled-screen guide-count analysis). Each skill provides a structured workflow, validation helpers, and reference documentation to ensure reproducible execution and result interpretation for these specific scientific domains.

(repo-wide) · high confidence

New scientific brainstorming skill with structured ideation workflows and validation scripts

This change introduces the \scientific-brainstorming\ skill, providing a structured, evidence-aware process for early-stage research ideation. The skill includes comprehensive documentation (\SKILL.md\ and reference guides) that enforces strict separation between idea generation, evidence assessment, and ethical/regulatory review, while mandating independent generation before group discussion to reduce anchoring bias. It also bundles Python CLI tools (\session\_scaffold.py\, \evaluate\_matrix.py\) and shared validation logic (\\_common.py\) to help users create deterministic session records, validate inputs, and perform transparent, weighted idea evaluation with sensitivity analysis. The skill is designed to be host-compatible and makes no external network or LLM calls by default.

skills/scientific-brainstorming · high confidence

New scientific presentation assets and templates

Added Beamer LaTeX templates for conference talks, dissertation defenses, and seminars, along with a PowerPoint design guide and presentation timing guidelines to support the creation of professional scientific slides.

skills/scientific-slides/assets · high confidence

New scientific slides scripts for schematic, slide image, PDF, and validation workflows

Added five new Python scripts to the scientific-slides skill: generate\_schematic.py and generate\_slide\_image.py use the Nano Banana 2 model via OpenRouter (reviewed by Gemini 3.7 Flash) to create scientific diagrams and presentation visuals with smart iteration and document-type-aware quality thresholds; pdf\_to\_images.py converts PDF presentations to images using PyMuPDF; slides\_to\_pdf.py combines slide images into a single PDF using Pillow; and validate\_presentation.py checks presentations (PDF, PPTX, LaTeX) for issues like slide count vs. duration, file size, and aspect ratio. All scripts resolve API keys securely and pass them to subprocesses via a minimal environment.

skills/scientific-slides/scripts · high confidence

New scikit-learn skill with v1.9.1 documentation and examples

Added a new scikit-learn skill (version 1.5) targeting scikit-learn 1.9.1 and Python 3.11+. The skill provides comprehensive reference documentation covering supervised and unsupervised learning, model evaluation, preprocessing, and pipelines, along with executable example scripts for classification and clustering workflows.

skills/scikit-learn · high confidence

New scikit-survival skill for survival analysis workflows

Added a new skill for scikit-survival 0.28.0 that provides documentation, reference guides, and local CLI tools for building and evaluating survival models. The skill covers right-censored outcomes, competing risks, and various model types including Cox PH, Coxnet, IPC ridge, survival forests, and SVMs. It includes detailed guidance on leakage-safe preprocessing, proper outcome construction, and censoring-aware evaluation metrics like concordance indices and Brier scores. The bundled scripts enable local, network-free execution of survival analysis workflows with input validation and bounded resource usage.

skills/scikit-survival · high confidence

New statistical analysis skill with comprehensive documentation and assumption-checking scripts

The \skills/statistical-analysis\ directory now contains a new skill for conducting guided statistical analyses, including hypothesis tests (t-tests, ANOVA, chi-square), regression, correlation, and Bayesian methods. The skill includes a main \SKILL.md\ guide, reference documents for assumptions, effect sizes, reporting standards, and test selection, along with a Python script (\scripts/assumption\_checks.py\) for automated assumption diagnostics. It targets specific versions of scientific Python libraries (Pingouin 0.7.0, SciPy 1.18.1, statsmodels 0.15.0, PyMC 6.3.2, ArviZ 1.3.0) and requires Python 3.12+.

skills/statistical-analysis · high confidence

New uncertainty-and-units skill for scientific calculation validation

A new 'uncertainty-and-units' skill has been added to the skills directory, providing a comprehensive framework for tracking physical units and propagating measurement uncertainty in scientific calculations. This skill bundles local command-line tools for uncertainty propagation, budget building, unit conversion, and plausibility checking, and includes extensive reference documentation covering the GUM methodology, pint library recipes, domain-specific conversion traps, and magnitude-based sanity checks. It is designed to help users avoid common errors such as silent unit stripping, incorrect temperature arithmetic, and invalid linearization assumptions by enforcing strict unit attachment and providing automated checks for dimensional consistency and physical plausibility.

skills/uncertainty-and-units · high confidence

New validation scripts for clinical decision support artifacts

Added seven new Python scripts in the clinical-decision-support skills to validate and generate reports for research-only CDS artifacts. These tools enforce safety and governance constraints by checking input documents for required metadata, prohibited clinical logic (e.g., diagnosis, dosing), and safe data boundaries (aggregate/synthetic only). Specific scripts include: \cohort\_table\_generator\ for creating disclosure-controlled tables with suppression rules; \decision\_logic\_traceability\ for validating governance decision-logic graphs; \deidentification\_checklist\ for verifying HIPAA Safe Harbor or expert determination documentation; \evidence\_profile\_check\ for validating GRADE evidence profiles; \model\_biomarker\_evaluation\ for generating aggregate model validation reports; \survival\_plan\_validator\ for checking estimand-led survival analysis plans; and \validate\_cds\_artifact\ for ensuring overall artifact compliance with research-only intended use and governance standards.

skills/clinical-decision-support/scripts · high confidence

Paper Lookup skill updated to v2.4 with companion literature APIs

The paper-lookup skill has been updated to version 2.4, introducing companion literature APIs and enhanced documentation. The skill now searches 18 scholarly APIs—including PubMed, PMC, Europe PMC, bioRxiv, medRxiv, arXiv, OpenAlex, Crossref, Semantic Scholar, CORE, Unpaywall, OpenCitations, PubTator3, Zenodo, Figshare, ROR, BioStudies, and DOAJ—returning results with reproducible provenance. The update includes detailed reference files for each database, covering endpoint contracts, parameter requirements, response shapes, and specific failure modes (such as silent HTTP 200 errors). It also provides guidance on identifier formats, API key usage, rate limiting, and pagination strategies to ensure reliable and auditable literature retrieval.

skills/paper-lookup · high confidence

Repository rebranded to Scientific Agent Skills with open standard support

The project has been renamed from 'claude-scientific-skills' to 'Scientific Agent Skills' and updated to version 2.72.0. It now supports the open Agent Skills and Agent Plugins standards, enabling compatibility with multiple AI agent hosts including Cursor, Claude Code, Codex, and Google Antigravity, rather than being limited to Claude. The repository includes a new \plugin.json\ manifest, a \CITATION.cff\ file for academic referencing, and a comprehensive \AGENTS.md\ guide for contributors. Additionally, a new automated security scanning workflow has been introduced to scan skills for vulnerabilities.

(repo-wide) · high confidence

Scientific visualization skill adds publication-ready style presets and publisher profiles

The scientific-visualization skill now includes curated Matplotlib style presets (nature, publication, presentation) and a machine-readable publisher profiles JSON to help users generate figures that align with major journal requirements. It bundles accessible color palettes (Okabe-Ito, Paul Tol) with contrast metadata, and provides reference documentation on color accessibility, journal-specific figure guidelines, and current Matplotlib/Seaborn/Plotly usage patterns. These assets support creating and auditing truthful, accessible, publication-ready scientific figures with Matplotlib, Seaborn, or Plotly, including explicit export provenance and integrity checks.

skills/scientific-visualization · high confidence

Venue Templates skill adds 2026-compliant LaTeX scaffolds and writing examples

The venue-templates skill now includes a comprehensive set of bundled assets for preparing academic and grant documents. This update adds official-style LaTeX templates for Elsevier journals (elsarticle v2.1 with numbered and author-year citation styles), specific scaffolds for NSF and NIH grant proposals, and a poster template. It also introduces illustrative writing examples for Cell, medical journals (NEJM, JAMA, BMJ), Nature/Science, and ML conferences (NeurIPS/ICML), all reviewed for 2026 guidelines. The skill's core documentation enforces a verification-first workflow, requiring users to check current venue rules before using these scaffolds.

skills/venue-templates · high confidence

Behavioural changes

Citation management skill rewritten with robust BibTeX parsing and multi-source metadata extraction

The citation-management skill has been replaced with a new suite of Python scripts that provide a shared, brace-depth-aware BibTeX parser to prevent title corruption and unbalanced braces in generated \.bib\ files. Users can now extract citation metadata from DOIs, PMIDs, arXiv IDs, and URLs via CrossRef, PubMed, and OpenAlex APIs, or search Google Scholar directly. The skill also includes tools to format, clean, and deduplicate existing BibTeX files, and to validate entries against required/recommended fields and venue-specific reference counts.

skills/citation-management/scripts · high confidence

DNAnexus integration skill updated to v2.3 with security and API corrections

The DNAnexus integration skill has been updated to version 2.3, aligning with dxpy 0.415.0 and dxCompiler 2.18.0. This release introduces stricter security guidance, explicitly prohibiting the logging or exposure of \DX\_SECURITY\_CONTEXT\ and API tokens, and correcting legacy patterns such as the removal of the deprecated \runSpec.systemRequirements\ field in favor of \regionalOptions\. It also updates Python SDK usage to reflect current \dxpy\ signatures (e.g., using \dxpy.open\_dxfile()\ instead of \DXFile.open\_file()\) and enforces safer data handling practices like using \name\_mode="glob"\ for searches and passing subprocess arguments as arrays to prevent injection.

skills/dnanexus-integration · high confidence

Fictiv skill documentation refreshed with September 2026 UI snapshot and safety guardrails

The Fictiv skill's reference documentation has been updated to reflect the authenticated web application state as of September 2026, including current material lists, process capabilities, and UI navigation paths. The update introduces stricter safety protocols for financial actions, specifically mandating an explicit user approval gate before placing orders and prohibiting the agent from handling raw payment card data (relying instead on Stripe iframes or password managers). It also clarifies export control constraints, explicitly stating that ITAR items are not supported in self-service workflows and require off-platform review.

skills/fictiv · medium confidence

FluidSim skill updated to version 1.4 with FluidSim 0.9.0 API alignment and safety tooling

The FluidSim skill has been updated to version 1.4, aligning its documentation and workflows with the FluidSim 0.9.0 API. This includes correcting parameter names (e.g., \cfl\_coef\ instead of \CFL\), updating solver registry keys (e.g., \plate2d\ replacing \fvk\), and refining initial condition and forcing configurations. The skill now bundles a set of Python scripts (\solver\_config\_validator\, \grid\_resource\_estimator\, \simulation\_dry\_run\, \output\_inventory\, \budget\_summary\) that enforce bounded, safe numerical workflows with strict resource limits and validation profiles for Cartesian solvers.

skills/fluidsim · high confidence

Geniml skill updated to v0.8.4 with local safety scripts and reference documentation

The Geniml skill now targets geniml 0.8.4 and gtars 0.10.0, with updated API maps for Region2Vec, scEmbed, and consensus-universe workflows. New dependency-free local scripts (bed\_validator, consensus\_plan, corpus\_auditor, tokenizer\_compatibility, embedding\_plan) validate inputs, plan executions, and enforce safety bounds without executing external commands or making network requests. Reference documents detail known compatibility issues, such as Gtars 0.10.0 breaking scEmbed's tokenize\_anndata and ScEmbed.encode, and BEDspace's reliance on the archived StarSpace executable.

skills/geniml · high confidence

GeoMaster skill documentation and reference updates

The GeoMaster skill documentation has been refreshed with new reference guides covering advanced GIS topics (3D geometry, viewshed, trajectories, topology), big data and cloud geospatial computing (Dask-GeoPandas, lazy rasters, cloud platforms, GPU paths, storage formats like COG and Zarr), and detailed API contracts for data sources (STAC, CDS, OSMnx, geocoding). The main SKILL.md has been updated to version 1.5, specifying tested library versions (e.g., GeoPandas 1.2.0, Rasterio 1.5.2) and installation instructions using uv, while also clarifying workflow best practices for CRS handling, raster alignment, and error prevention in spatial analysis.

skills/geomaster · high confidence

Image generation skill now uses the OpenRouter Image API

The generate-image skill has been updated to route image generation and editing requests through the OpenRouter Image API (POST /api/v1/images) instead of its previous provider. This change introduces support for a broader catalogue of models—including Gemini, Seedream, Recraft, GPT-Image, and Riverflow—while enforcing stricter parameter validation via preflight discovery to prevent billing for invalid requests. Users must now provide an OpenRouter API key (via OPENROUTER\_API\_KEY or --api-key) and can leverage new capabilities such as transparent backgrounds, vector SVG output, and multi-image generation, with detailed model-specific constraints documented in the updated skill metadata and reference files.

skills/generate-image · high confidence

LabArchives integration skill updated to current API specifications and security practices

The LabArchives integration skill has been refreshed to align with the latest official LabArchives ELN and Inventory API v1 documentation. The skill now explicitly distinguishes between the legacy ELN API (using query parameters) and the Inventory API v1 (using headers), enforcing correct regional endpoint selection and HMAC-SHA-512 signature construction for both. Security and operational safeguards have been strengthened, including strict HTTPS enforcement, validation of request paths, limits on LA container inspection, and the removal of any reliance on \.env\ files in favor of explicit environment variables. The skill also clarifies that advertised product integrations (e.g., Jupyter, REDCap) are file-based or UI workflows and do not imply generic API access.

skills/labarchive-integration · high confidence

LatchBio Integration skill updated to SDK 2.77.1 with MCP support and corrected API references

The LatchBio integration skill has been refreshed to target Latch SDK 2.77.1, introducing documentation for the new Latch MCP (Model Context Protocol) server for AI-driven workflow discovery and execution. The skill now provides updated references for Latch Data operations, specifically distinguishing between the modern \LPath\ API and the established \LatchFile\/\LatchDir\ types, and corrects Registry API usage to use the new object- and transaction-based model. It also clarifies Nextflow and Snakemake packaging paths, noting the deprecation of legacy Snakemake flags in favor of the v2 tutorial track, and updates resource configuration examples to reflect current scheduler requests for CPU, GPU, and custom task shapes.

(repo-wide) · high confidence

Open Notebook v1.14.0 API client scripts

The Open Notebook example scripts have been updated to target the v1.14.0 API, introducing new capabilities for managing notebooks, ingesting sources, and interacting with chat sessions. Users can now create, list, update, and delete notebooks, as well as link or unlink sources. Source ingestion supports adding URLs, text, and file uploads with configurable async processing and embedding options, including a robust polling mechanism to wait for processing completion. Chat interactions now allow creating sessions, building explicit context from specific sources and notes, sending messages, retrieving history, and searching the knowledge base. The scripts also include helpers for deleting sources and retrieving source insights, all utilizing a shared authenticated HTTP client.

skills/open-notebook/scripts · high confidence

PathML skill documentation and tooling updated for version 3.0.8

The PathML skill has been updated to align with the PathML 3.0.8 stable release. The documentation now reflects the correct API usage, such as using \slide.run()\ instead of \Pipeline.run()\, and clarifies that \SlideData.from\_slide()\ is not a stable method. It also corrects the handling of multiplex data, noting that \CollapseRunsCODEX\ and \CollapseRunsVectra\ do not perform spectral unmixing or registration. Security guidance has been strengthened to warn against using untrusted \.pt\ checkpoints and to clarify that \SegmentMIFRemote\ downloads models from Hugging Face at construction time. Additionally, new CLI scripts for manifest validation and image QC have been added to support bounded, local research workflows.

skills/pathml · high confidence

PyLabRobot skill updated to version 1.5 with verified 0.2.2 API references and safety workflows

The PyLabRobot skill has been updated to version 1.5, aligning its documentation and offline tooling with the PyLabRobot 0.2.2 release. This update introduces a strict safety boundary that prevents automatic physical device execution, requiring explicit human confirmation for any live hardware interaction. It provides verified API references for liquid handling, analytical equipment, and material handling, correcting deprecated import names and method signatures. Additionally, the skill now includes a JSON schema for protocol manifests and bundled CLI scripts for offline manifest validation, deck geometry checking, and backend inspection, ensuring that planning and simulation remain isolated from physical execution.

skills/pylabrobot · high confidence

QuTiP skill updated to version 5.3.1 with new CLI tools and documentation

The QuTiP skill now targets QuTiP 5.3.1, requiring Python 3.11+ and updating all solver examples and API references to match the new keyword-only argument signatures (e.g., \options\ as a dictionary instead of the old \Options\ object). This release introduces a suite of command-line tools for bounded, reproducible quantum simulations, including a model validator (\qobj\_model\_validator.py\), a convergence sweep runner (\convergence\_sweep.py\), and a result auditor (\result\_audit.py\), all of which enforce strict input limits and JSON-based I/O. Comprehensive documentation has been added to \SKILL.md\ and the \references/\ directory, covering core concepts, time evolution, advanced methods like Bloch-Redfield and HEOM, and visualization guidelines specific to QuTiP 5.3.

skills/qutip · high confidence

Redesigned PPTX poster generation with strict validation and safety controls

The PPTX poster generation skill has been redesigned to enforce strict safety, reproducibility, and quality standards. The new scripts validate poster manifests against a new schema (v2.0), ensuring inputs are bounded, local, and free of dangerous content (e.g., macros, OLE objects, symlinks). Generation now requires exact dependency versions (python-pptx 1.0.2, Pillow 12.3.0, lxml 6.1.3) to guarantee reproducible output. The toolchain includes dedicated checks for layout integrity (overlap, reading order, font size), color accessibility (WCAG 2.2 contrast ratios), and image quality (effective DPI, hidden metadata). Additionally, a new export plan script provides manual preflight instructions for conference and printer requirements, separating the automated generation from the final human-reviewed export steps.

skills/pptx-posters/scripts · high confidence

Redesigned scholar-evaluation skill with strict safety boundaries and local-only tooling

The scholar-evaluation skill has been redesigned to enforce a hard safety boundary prohibiting use in high-impact personnel decisions (hiring, promotion, tenure, admissions, funding, awards, discipline) and person ranking. The skill now provides a qualitative-first, evidence-traceable developmental review framework for scholarly works, supported by new local JSON/CSV templates for rubrics, evaluations, evidence manifests, process checklists, and ratings. All bundled Python scripts are strictly local, using only the standard library, with no network access, external model calls, credential handling, or subprocess execution, ensuring deterministic and secure operation.

skills/scholar-evaluation · high confidence

Redesigned scientific writing safeguards with fail-closed scaffolds and evidence registries

The scientific-writing skill now enforces a fail-closed workflow that prevents submission-ready formatting until all evidence is verified. It introduces structured registries (source manifest, claims CSV, consistency manifest, authorship JSON, and reporting coverage JSON) and a Markdown manuscript scaffold that keeps placeholders visible until human approval. Local Python 3.11+ CLI tools scaffold the workspace, select reporting guidelines, validate manifests, audit claims and references, check numeric/methods consistency, and lint for unresolved placeholders. The skill also updates reporting guideline coverage to include CONSORT 2025 and SPIRIT 2025 alongside existing standards, and clarifies confidentiality, authorship, and AI-use disclosure requirements aligned with ICMJE and COPE guidance.

skills/scientific-writing · high confidence

Redesigned treatment-plan skill with strict documentation-only scope and safety boundaries

The treatment-plans skill has been redesigned to enforce a strict documentation-only workflow, explicitly prohibiting clinical decision-making, diagnosis, or therapy selection. It now uses a new set of generic JSON templates (schema v2.0) for structured data entry and includes a suite of local, deterministic validation scripts that check for structural integrity, traceability, and privacy compliance without accessing external networks or models. The skill mandates visible 'DRAFT' notices, requires authorized clinician sign-off for any release, and implements hard safety boundaries to refuse any requests that cross into clinical advice or emergency triage.

skills/treatment-plans · high confidence

Research Lookup skill now targets 60 verified references with structured manuscript packets

The Research Lookup skill has been updated to compile a structured research packet targeting 60 verified, unique references by default, rather than a loose list of links. The new workflow uses Parallel Search and Extract to deduplicate and verify sources, producing artifacts including an evidence matrix, claim-to-source map, synthesis of consensus and gaps, and section briefs for Introduction, Methods, and Discussion. The skill retains backward compatibility with the existing ResearchLookup class and CLI flags (such as --batch, --json, and -o/--output), while introducing explicit backend routing for Parallel Research, Parallel Chat, and optional Perplexity via OpenRouter.

skills/research-lookup · high confidence

SimPy skill rewritten with bounded, safe CLI tools and strict input validation

The SimPy skill scripts have been replaced with a new set of bounded, deterministic command-line tools that enforce strict safety and input constraints. The new implementation pins the SimPy library to version 4.1.2 and rejects network URLs and symlinks for all input and output paths. It introduces specific CLI commands for running bounded queue scenarios, summarizing event traces and resource CSVs, running independent replications with Student-t confidence intervals, and validating simulation configurations without execution. All JSON inputs are validated against allowlists, numeric values are strictly bounded, and outputs are written atomically to prevent corruption. A specific fix ensures that non-UTF-8 resource CSV files are reported as unreadable rather than causing integer parsing errors.

skills/simpy/scripts · high confidence

Updated Gtars skill documentation and added local safety scripts

The Gtars skill documentation has been updated to reflect the current 0.10.0 Python and Rust APIs, including corrected import paths, verified method signatures, and clarified behavioral nuances for overlap, consensus, and coverage operations. To support secure usage of these native components, the skill now includes a suite of dependency-free local helper scripts (e.g., \artifact\_inspector.py\, \\_common.py\) that enforce strict input safety contracts, such as rejecting symlinks, URL schemes, and unbounded file sizes, ensuring that genomic data processing remains isolated and resource-controlled.

skills/gtars · high confidence

Updated NeuroKit2 skill documentation for version 0.2.13

The NeuroKit2 skill documentation has been refreshed to align with the stable NeuroKit2 0.2.13 release. The updated SKILL.md and reference guides now specify the correct Python 3.10+ compatibility, pinned dependencies (including pandas \<3.0.0), and accurate API schemas for core processing functions like ecg\_process, eda\_process, and bio\_process. The documentation also clarifies specific behavioral details, such as the default high-pass decomposition for EDA, the flat structure of info dictionaries, and the requirements for optional dependencies like MNE for EEG and EOG processing.

skills/neurokit2 · high confidence

Updated protocols.io integration with current API contracts and OpenClaw support

The protocols.io integration skill has been refreshed to align with the official protocols.io API reference as of September 2026. The skill now correctly distinguishes between mixed v3 and v4 endpoints, explicitly mapping operations like protocol retrieval and step management to their current v4 paths while retaining v3 for list and material endpoints. It introduces support for the official protocols.io MCP server, documenting its read-only tools and public-content scope. The integration also adds support for OpenClaw via the \PROTOCOLS\_IO\_ACCESS\_TOKEN\ environment variable and includes a new local JSON schema for validating protocol snapshots offline. All documentation and reference files have been updated to reflect the latest API contracts, authentication flows, and safety guidelines for handling untrusted content.

skills/protocolsio-integration · high confidence

deepTools skill updated to version 2.0 targeting deepTools 4.0.0

The deepTools skill has been updated to version 2.0, aligning with the upstream deepTools 4.0.0 release. This update introduces a comprehensive set of documentation files (SKILL.md, quick reference, and detailed references) and utility scripts (file validation and workflow generation) to guide users through ChIP-seq, RNA-seq, and ATAC-seq analyses. Key behavioral changes for users include the removal of legacy options like \--ignoreDuplicates\ and \--exactScaling\, the replacement of SES scaling with \readCount\ or \None\ in \bamCompare\, and the clarification that BPM normalization now numerically equals CPM in the Rust-backed implementation. The skill also provides corrected effective genome size values specific to the 4.0.0 tagged source and highlights known limitations, such as the behavior of \--scaleFactorsMethod\ and the need for careful handling of zero-valued bins in bigWig comparisons.

skills/deeptools · high confidence

Test coverage

New test infrastructure and validation for the 13C metabolic flux skill

Added a comprehensive test suite for the 13C metabolic flux skill, including regression tests for numerical accuracy (e.g., flux recovery, mass balance, and identifiability assessments) and an agent evaluation benchmark. Introduced a shared test contract framework (\tests/\_contract\) that enforces structural compliance, CLI help availability, and secure environment handling across all skills. Added meta-tests to validate plugin manifests against the Agent Plugins 1.0.0 schema and ensure documentation integrity.

tests · high confidence

Dependencies

Update project dependencies and add new skill requirements

The project has been updated to version 2.72.0 with new dependencies including cisco-ai-skill-scanner, pytest, and python-dotenv. Additionally, new skill requirements have been added for 13c-metabolic-flux (mfapy, numpy, scipy, nlopt), opentrons-integration (opentrons 10.0.0 for flex, 9.0.0 for ot2), and primer-design (primer3-py).

(dependencies) · high confidence

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

Score

  • CAI 59 → 53 (-6.5)
  • Rubric changed (rubric-2026.11.1 → rubric-2026.11.4) — scores are not directly comparable.

Lenses

  • Code Health 77 → 75 (-2.0)
  • Architecture 88 → 96 (+7.8)
  • Maturity 81 → 61 (-20.2)
  • Readiness 36 → 34 (-2.4)
  • Security 83 → 77 (-6.6)

Resolved (101)

  • Duplicated block (10 lines × 2) (skills/get-available-resources/scripts/detect_resources.py)
  • Duplicated block (10 lines × 3) (skills/pptx-posters/scripts/check_palette.py)
  • Duplicated block (10–12 lines × 2) (skills/clinical-decision-support/scripts/decision_logic_traceability.py)
  • Duplicated block (10–12 lines × 2) (skills/market-research-reports/scripts/validate_competitor_matrix.py)
  • Duplicated block (10–24 lines × 2) (skills/lab-hardware-cad/scripts/_common.py)
  • Duplicated block (11 lines × 2) (skills/pathml/scripts/slide_manifest.py)
  • Duplicated block (11 lines × 3) (skills/clinical-decision-support/scripts/evidence_profile_check.py)
  • Duplicated block (12 lines × 2) (skills/clinical-decision-support/scripts/model_biomarker_evaluation.py)
  • Duplicated block (12 lines × 2) (skills/neurokit2/scripts/ecg_hrv_pipeline.py)
  • Duplicated block (12 lines × 2) (skills/scientific-writing/scripts/check_references.py)
  • Duplicated block (12–14 lines × 3) (skills/geniml/scripts/embedding_plan.py)
  • Duplicated block (13 lines × 2) (skills/hypogenic/scripts/evaluate_local.py)
  • Duplicated block (13–15 lines × 2) (skills/scholar-evaluation/scripts/calculate_scores.py)
  • Duplicated block (14 lines × 2) (skills/iso-standards-readiness/scripts/audit_document_records.py)
  • Duplicated block (14 lines × 2) (skills/scientific-brainstorming/scripts/validate_register.py)
  • Duplicated block (14 lines × 3) (skills/clinical-reports/scripts/check_deidentification.py)
  • Duplicated block (14–16 lines × 2) (skills/geopandas/scripts/export_plan.py)
  • Duplicated block (14–17 lines × 2) (skills/scientific-brainstorming/scripts/validate_register.py)
  • Duplicated block (14–24 lines × 10) (skills/hypothesis-generation/scripts/check_falsification_controls.py)
  • Duplicated block (15 lines × 2) (skills/clinical-decision-support/scripts/deidentification_checklist.py)
  • …and 81 more

New (129)

  • Duplicated block (14 lines × 2) (skills/iso-standards-readiness/scripts/audit_document_records.py)
  • Duplicated block (21–22 lines × 2) (skills/opentrons-integration/scripts/pcr_setup_template.py)
  • Duplicated block (8 lines × 3) (skills/iso-standards-readiness/scripts/audit_document_records.py)
  • Edited copies of one member (3 members, 10+ corresponding lines) (skills/pathml/scripts/_common.py)
  • Edited copies of one member (3 members, 11+ corresponding lines) (skills/clinical-reports/scripts/_common.py)
  • Edited copies of one member (3 members, 12+ corresponding lines) (skills/simpy/scripts/replication_runner.py)
  • Edited copies of one member (3 members, 13+ corresponding lines) (skills/pptx-posters/scripts/check_palette.py)
  • Edited copies of one member (3 members, 13+ corresponding lines) (skills/protocolsio-integration/scripts/plan_write_request.py)
  • Edited copies of one member (3 members, 14+ corresponding lines) (skills/geniml/scripts/_common.py)
  • Edited copies of one member (3 members, 17+ corresponding lines) (skills/iso-standards-readiness/scripts/check_traceability.py)
  • Edited copies of one member (3 members, 20+ corresponding lines) (skills/pytorch-lightning/scripts/template_lightning_module.py)
  • Edited copies of one member (3 members, 22+ corresponding lines) (skills/scikit-survival/scripts/_common.py)
  • Edited copies of one member (3 members, 28+ corresponding lines) (skills/exploratory-data-analysis/scripts/_common.py)
  • Edited copies of one member (3 members, 32+ corresponding lines) (skills/exploratory-data-analysis/scripts/_common.py)
  • Edited copies of one member (3 members, 34+ corresponding lines) (skills/neurokit2/scripts/ecg_hrv_pipeline.py)
  • Edited copies of one member (3 members, 4+ corresponding lines) (skills/qutip/scripts/convergence_sweep.py)
  • Edited copies of one member (3 members, 5+ corresponding lines) (skills/bulk-rnaseq/scripts/build_counts_matrix.py)
  • Edited copies of one member (3 members, 5+ corresponding lines) (skills/hypogenic/scripts/_common.py)
  • Edited copies of one member (3 members, 6+ corresponding lines) (skills/exploratory-data-analysis/scripts/_tabular.py)
  • Edited copies of one member (3 members, 6+ corresponding lines) (skills/matlab/scripts/plan_python_compatibility.py)
  • …and 109 more

Changes since last survey

  • 1 commits — 1 feature/other, 0 fixes

By area

  • docs/security-report.json — 1 commit

Notable commits

  • change: chore: update security scan report [skip ci]

Architecture

  • Unchanged — 0 containers · 1 contexts · 0 edges

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

K-Dense-AI/scientific-agent-skills was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 15 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit 2d2065fb4520fa3f3f2bcf61cff26ec8b7eba4b5 — the exact code this score is about.
  • Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.73+preprod-52de90b74216.