Computer-science-claude-skills

Use when designing or implementing tool-using agents for research or software workflows and you need an explicit architecture, tool contract, prompt strategy, and evaluation plan.

Use when designing an evaluation plan for a CS, ML, or agent project and you need to choose datasets, baselines, metrics, ablations, compute budgets, and reporting rules.

Manage citations and BibTeX for computer science papers, theses, surveys, rebuttals, and project reports. Use when verifying DOI or arXiv metadata, cleaning `.bib` files, deduplicating references, formatting venue-ready citations, or checking that paper, code, and artifact references are consistent.

Use when reproducing a paper, repo, benchmark, or reported result and you need an explicit plan for environment setup, execution, discrepancy logging, and final reproduction status.

Use when turning an existing codebase, experiment set, or system into a paper plan and you need contribution framing, evidence mapping, figure planning, and missing-experiment detection.

Use when running iterative experiments and you need a disciplined way to log configs, seeds, environments, metrics, failures, and comparison summaries.

Explore and summarize computer science research data, experiment outputs, benchmark tables, logs, embeddings, predictions, and dataset files. Use when inspecting CSV, JSON, Parquet, NPY, NPZ, HDF5, TXT logs, or similar files to understand structure, quality, anomalies, and next analysis steps before modeling, benchmarking, or paper writing.

Detect available CPU, GPU, memory, and disk resources before compute-heavy CS work. Use when planning model training, large-scale evaluation, embedding generation, experiment sweeps, dataset processing, log analysis, or reproduction runs where hardware limits affect the workflow.

Use when preparing a project for public GitHub release and you need to verify docs, licenses, notices, examples, repository hygiene, and release readiness.

Create computer science research posters in LaTeX for conferences, thesis defenses, demos, and lab presentations. Use when converting a CS paper, benchmark result, system design, or project summary into a poster with strong layout, readable figures, and concise technical messaging.

Review and synthesize computer science literature for related-work sections, surveys, benchmark positioning, and paper comparison. Use when mapping a CS topic, building a reading list, tracing citation chains, comparing baselines, summarizing papers, or identifying research gaps in machine learning, systems, software engineering, HCI, security, theory, or adjacent computing areas.

Write markdown-first technical documents with Mermaid diagrams for CS research, engineering, and agent workflows. Use when creating design docs, research notes, experiment plans, repo documentation, issue reports, architecture diagrams, workflow descriptions, or text-native visual explanations that should diff cleanly in Git.

Convert papers, slide decks, reports, office files, and mixed-format project materials into Markdown for CS research and engineering workflows. Use when turning PDFs, PPTX, DOCX, HTML, JSON, CSV, or similar inputs into LLM-friendly text for literature review, repo cleanup, release prep, or paper-to-code analysis.

Create static, highly customized figures for CS papers, benchmark reports, experiment analysis, and technical documentation. Use when default plotting tools are too limiting, when multi-panel layouts need precise control, or when exporting clean PNG, PDF, or SVG figures for research outputs.

Build and analyze graphs in Python for CS research and engineering tasks. Use when working with citation graphs, dependency graphs, call graphs, knowledge graphs, agent tool graphs, communication networks, or any problem involving nodes, edges, graph algorithms, or network structure.

Use when implementing a paper, method, or system from a publication and you need to translate claims, methods, and diagrams into an executable implementation plan.

Review computer science papers, rebuttals, and artifact submissions for conferences, journals, and workshops. Use when evaluating novelty, technical correctness, empirical rigor, reproducibility, clarity, significance, or when drafting a structured review, meta-review, or author questions.

Use when a research or prototype repository has become messy and you need a disciplined cleanup plan for structure, naming, docs, unused files, configs, and reproducibility hygiene.

Use when starting or reframing a computer science research project and you need to turn a vague idea into a scoped plan with research questions, milestones, risks, and next actions.

Use when responding to paper reviews, rebuttal comments, or revision requests and you need a structured issue list, action plan, and concise response draft.

Score and assess CS papers, proposals, literature reviews, and benchmark plans using a structured evaluation rubric. Use when a quantitative or semi-structured assessment is needed for research quality, novelty, technical soundness, empirical rigor, reproducibility, clarity, and submission readiness.

Critically examine claims, assumptions, evidence, and argument quality in computer science papers, experiments, benchmark plans, and technical proposals. Use when stress-testing a research idea, detecting confounders, checking causal leaps, or identifying missing controls before writing, reviewing, or implementing.

Create polished technical diagrams for CS papers, slides, posters, demos, and repo documentation. Use when a Mermaid diagram is not enough and the task needs a cleaner architecture figure, system overview, pipeline graphic, benchmark setup diagram, or agent workflow illustration.

Build and revise computer science research slides for conference talks, paper presentations, defenses, group meetings, demos, and project updates. Use when turning a paper, benchmark, experiment, system design, or research plan into clear slides with strong structure, visual hierarchy, and timing discipline.

Orchestrate publication-ready figures for CS papers, posters, slides, and benchmark reports. Use when assembling multi-panel result figures, harmonizing styles across plots, preparing export settings, or turning experiment outputs into clear visual evidence for submissions and technical communication.

Draft and revise computer science papers, theses, reports, and technical research documents. Use when writing or rewriting abstracts, introductions, methods, experiments, limitations, artifact sections, related work, or when converting notes, code, and results into publishable CS prose.

Choose and report statistical analyses for CS experiments, benchmarks, user studies, and offline evaluations. Use when selecting tests, checking assumptions, comparing multiple runs or seeds, estimating uncertainty, reporting effect sizes, or deciding whether differences in results are credible.

Fit statistical models with detailed inference and diagnostics for CS experiments, telemetry, time series, and structured evaluations. Use when you need OLS, GLM, mixed models, logistic regression, count models, or time-series analysis with coefficient tables, residual checks, and interpretable model summaries.

Use symbolic math in Python for CS research and algorithm work. Use when deriving formulas, simplifying objectives, solving recurrences or equations, manipulating matrices symbolically, generating exact expressions, or turning mathematical derivations into executable code.

Build graph neural network workflows with PyTorch Geometric for CS research. Use when training GNNs for citation networks, knowledge graphs, recommender graphs, code graphs, program analysis, heterogeneous graphs, node or graph classification, or link prediction with scalable mini-batching and graph transforms.

Work with transformer models for CS research and development. Use when loading, fine-tuning, evaluating, or serving language, vision, audio, multimodal, embedding, or LLM-style models for tasks such as generation, classification, retrieval, instruction tuning, benchmarking, or agent backends.

Reduce and visualize high-dimensional representations for CS experiments. Use when exploring embeddings, hidden states, feature vectors, retrieval spaces, clustering structure, error slices, or representation drift in model outputs and benchmark data.

Process very large tabular data for CS experiments and engineering logs without loading everything into RAM. Use when working with massive result tables, telemetry, predictions, feature exports, ranking data, or other large CSV, HDF5, Arrow, or Parquet files that need fast lazy aggregation and filtering.

Adapt computer science papers, posters, rebuttals, and submission packages to venue-specific templates and style expectations. Use when targeting ACM, IEEE, USENIX, NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, EMNLP, ACL, NAACL, AAAI, IJCAI, KDD, WWW, CHI, UIST, ICSE, FSE, ASE, SOSP, OSDI, NSDI, or similar CS venues.