Claude Skills
The open format is called Agent Skills and works in Claude Code, Codex, Cursor and other agents — most people know it as Claude Skills.
Every Agent Skill we could find on GitHub, deduplicated by content. 79 437 files from 1 744 authors, of which 61 785 are unique — the rest is the same skill repackaged into someone else's repository. For each one: what it weighs in tokens, whether it ships runnable scripts, and which MCP servers it needs.
Use as the routing layer for Python development tasks — matches task descriptions against trigger lists and activates specialist skills before starting work. Covers Typer, Rich, Textual, FastMCP/MCP, ty type checker, uv, Hatchling, TOML editing, pre-commit/prek, async Python, PyPI packaging, complex linting, and technical debt modernization.
LAST RESORT for stdlib-only Python 3.11+ scripts in CONFIRMED restricted environments (airgapped systems, no uv, no internet access). Creates portable dependency-free scripts using argparse, logging, config management (JSON/TOML/INI), and cross-platform patterns. Use ONLY when environment restrictions prevent Typer+Rich with PEP 723. Triggers on "stdlib-only script", "no dependencies", "airgapped", "restricted environment", "portable script no network". For standard CLI development, use python-cli-architect with Typer+Rich instead.
Progressive Python quality improvement with static analysis, type refinement, modernization planning, plan review, and test-driven implementation. Use when addressing technical debt, eliminating Any types, applying modern Python patterns, or refactoring for better design.
Use when encountering failing tests, diagnosing test errors, or establishing a systematic approach to test failure investigation. Activates on "test failure analysis", "debugging tests", or "why tests fail" requests. Establishes the mindset that treats test failures as valuable diagnostic signals requiring root-cause investigation — not automatic code fixes or test dismissal.
Use when building Textual TUI apps — creating widgets, laying out screens, handling events, managing reactivity, testing with Pilot, or running background workers. Covers App lifecycle, CSS styling, screen stack, reactive attributes, custom messages, actions, bindings, and the Worker API.
Use when reading or writing pyproject.toml or .toml config files in Python, editing TOML while preserving comments and formatting, designing configuration file formats for Python tools, working with tomlkit or tomllib, or implementing atomic config file updates.
Use when working with ty — running Python type checks, configuring ty.toml or pyproject.toml, suppressing diagnostics, interpreting error codes, targeting Python versions, or integrating ty with editors and CI. Covers CLI flags, configuration schema, rule severity, suppression comments, environment discovery, module resolution, and all installation methods.
Use when building Typer/Rich CLI applications or reviewing existing CLI code for correctness. Activates on requests involving Rich table rendering, console output handling, testing Rich-formatted output, or Typer command wiring. Prevents common AI mistakes — Rich table wrapping in non-TTY contexts, incorrect stderr/stdout separation, and integration pitfalls. Load alongside the typer and rich API reference skills.
Use when building CLI applications with Typer — creating commands, defining arguments and options, composing subcommands, testing with CliRunner, or using advanced features like callbacks and autocompletion. Covers app creation, parameter types, subcommand composition, testing patterns, and output utilities.
Use when working with Astral's uv — managing Python project dependencies, creating PEP 723 scripts, installing tools, managing Python versions, configuring package indexes, or migrating from pip/poetry. Covers project initialization, virtual environments, workspace configuration, and CI/CD integration.
Scan Claude Code session transcripts to find the strongest user reactions to assistant instruction-following failures, reconstruct the triggering assistant output, and render a shareable terminal-style PNG artifact. Use when you want to surface and share a moment where the assistant completely missed what was asked — captures what they were doing, what Claude said, and how the user reacted. Triggers on: "rtfp", "read the fucking prompt", "find my worst AI moment", "make a rage screenshot from this session".
Use when debugging software, investigating incidents, diagnosing flaky tests, or analyzing performance regressions — enforces structured observation recording with evidence IDs, causality validation, and verification gates to prevent correlation-causation pollution. Use when an agent might otherwise summarize or speculate instead of reporting observed evidence.
Design and run controlled experiments using the experiment-registry MCP server — domain-agnostic, pluggable, mechanically enforced. Use when you need evidence that a change actually improves behaviour.
Use when facing unknowns, debugging without a clear cause, or making architecture decisions — enforces hypothesis-driven scientific reasoning through observation, hypothesis formulation, prediction, experiment design, and evidence-based conclusion. Use when previous attempts have failed or the problem space involves uncertainty.
Rules for orchestrators handling sub-agent results without lossy re-summarization. Activates when receiving agent output, relaying research results, passing data between agents, or combining agent findings. Prevents the failure mode where 'not found' becomes 'doesn't exist' and counts get dropped.
Summarize files by reading content, extracting key passages, and applying type-specific strategies. Activates on summarize this file, what's in this file, describe this codebase, file summary, analyze this file, tl;dr this file, what does this code do, explain this config, break down this script. Routes to strategies for code, config, data, documentation, markup, and binary files based on extension and word count.
Describe images and screenshots by viewing content with multimodal Read tool and documenting only visible elements. Activates on what's in this image, describe this screenshot, summarize this diagram, image summary, what does this screenshot show, explain this diagram, break down this chart. Handles UI screenshots, architecture diagrams, charts, photos, code screenshots, and terminal output. Never infers from filenames.
Combines multiple summaries into coherent synthesis. Activates when user requests combine these summaries, synthesize results, merge findings, or multi-source synthesis. This is the reduce step after individual sources have been summarized. Enforces deduplication, conflict surfacing, confidence scoring across sources, and attribution to all sources.
Route summarization requests to the correct methodology and enforce fidelity rules. Activates on summarize, tl;dr, give me the highlights, what's important in this, break down this, what does this code do, explain this file, describe this image, read and summarize. Routes files, URLs, images, and multi-source content to type-specific strategies. Enforces anti-hallucination rules — read before summarizing, extract before abstracting, preserve counts, distinguish absence from nonexistence, prevent lossy re-summarization chains.
Summarize web content by fetching URLs, extracting key passages with quote-grounding, and producing structured output. Activates on summarize this URL, what does this page say, summarize this article, read and summarize, summarize the documentation at, tl;dr this link, give me the highlights of this page, what's important on this site. Routes to fetching strategy based on content type — documentation, articles, API references, READMEs. Reports partial accessibility explicitly.
Use when auditing docs vs code drift, syncing docs after changes, optimizing CLAUDE.md or SKILL.md for AI consumption, validating GLFM and Markdown, or summarizing files/URLs/images — routes each task to the correct specialist agent via /rwr:audit, /rwr:optimize, or /rwr:author
Converts user-facing documentation (how-to guides, tutorials, API references, examples) in any format — Markdown, PDF, DOCX, PPTX, XLSX, AsciiDoc, RST, HTML, Jupyter notebooks, man pages, TOML/YAML/JSON configs, and plain text — into Claude Code skill directories with SKILL.md plus thematically grouped references/*.md files. Use when given a docs directory or mixed-format documentation to transform into an AI skill. Uses MCP file-reader server for binary formats.
Reference guide for the Twelve-Factor App methodology — 15 principles (12 original + 3 modern extensions) for building portable, resilient, cloud-native applications. Use when evaluating application architecture, designing cloud-native services, reviewing codebases for methodology compliance, advising on configuration, scaling, observability, security, and deployment patterns. Incorporates the 2025 open-source community evolution and cloud-native reinterpretations of each factor.
Enforce mandatory pre-action verification checkpoints to prevent pattern-matching from overriding explicit reasoning. Use this skill when about to execute implementation actions (Bash, Write, Edit) to verify hypothesis-action alignment. Blocks execution when hypothesis unverified or action targets different system than hypothesis identified. Critical for preventing cognitive dissonance where correct diagnosis leads to wrong implementation.
When an application needs to store config, data, cache, or state files. When designing where user-specific files should live. When code writes to ~/.appname or hardcoded home paths. When implementing cross-platform file storage with platformdirs.
Responsive HTML email template generation using MJML 4.x framework. Use when the user asks to create, generate, design, or build an email template — including welcome emails, promotional blasts, transactional templates, newsletters, or any responsive email. Also use when the user asks to compile MJML to HTML, work with or edit existing MJML templates, or troubleshoot email rendering issues across clients.
Swift iOS architecture guidance and playbooks for MVVM, MVI, TCA, Clean Architecture, VIPER, MVP, Coordinator, and Reactive patterns. Use when designing, implementing, refactoring, or reviewing the architecture of a SwiftUI or UIKit feature, module, or codebase.
Редполитика проекта на русском языке: короткое интервью, извлечение голоса из примеров и файл `.agents/redpolitika.md` для редакторских скиллов. Use this skill whenever the user asks to create, update, remember, formalize, or apply a project editorial policy, tone of voice, style guide, vocabulary, forbidden words, or reusable writing context. Триггеры: «настрой редполитику», «редполитика», «запомни мой стиль», «создай гайд по стилю», «tone of voice», «ToV», «настрой скилл под наш проект», «как нам писать», «зафиксируй голос», «добавь в редполитику», «обнови правила тона», «словарь проекта», «запрещённые слова», «пиши в нашем стиле», «сохрани правила редакции». Also suggest this skill when redaktura, statya, or ux-copy lacks project context and the user will keep working with the same product. Для редактуры готового текста см. skill redaktura; для статей с нуля — statya; для интерфейса — ux-copy.
Редактура текстов на русском языке по методу Людмилы Сарычевой — от смысла к словам, а не наоборот. Use this skill whenever the user asks to edit, review, improve, shorten, or polish any Russian-language text — articles, posts, emails, landing copy, docs. Триггеры: «отредачь», «отредактируй», «почисти текст», «сократи», «перепиши по-человечески», «убери канцелярит», «сделай проще», «причешите текст», «дай фидбек на текст», «разбери мой текст», «слишком водянисто», «звучит как робот». Also trigger when the user pastes a Russian draft and asks what's wrong with it or how to make it better. Для написания текста с нуля см. skill statya; для текстов интерфейса — ux-copy.
UX-тексты на русском языке: кнопки, формы, ошибки, пустые состояния, онбординг, подтверждения, уведомления, пуши и микрокопия. Use this skill whenever the user asks to write, rewrite, review, choose, or improve interface copy, product messages, UI strings, transactional emails, push notifications, toasts, snackbars, forms, empty states, onboarding, errors, confirmations, labels, tooltips, or button text. Триггеры: «текст для кнопки», «сообщение об ошибке», «текст в форму», «пустое состояние», «подпись к полю», «текст тоста», «снекбар», «онбординг», «пуш-уведомление», «микрокопия», «UX-текст», «что написать на экране…», «лейбл поля», «тултип», «подтверждение удаления», «письмо-уведомление». Для статей и постов см. skills statya и redaktura; для проектного словаря — redpolitika.
Написание русскоязычной статьи или лонгрида с нуля: бриф, фактура, структура, черновик и самопроверка по методу Людмилы Сарычевой. Use this skill whenever the user asks to write, outline, structure, draft, assemble, or turn notes/interviews/materials into a Russian article, newsletter essay, longread, case, guide, or expert text. Триггеры: «напиши статью», «собери лонгрид», «напиши текст для блога про…», «есть интервью с экспертом, сделай статью», «составь план статьи», «придумай структуру», «собери материал», «сделай черновик», «напиши колонку», «из интервью в статью», «план лонгрида», «текст с нуля», «помоги написать материал». Для коротких постов в соцсети см. skill post; для правки готового текста — redaktura; для текстов интерфейса — ux-copy; для настройки проекта — redpolitika.
Продающие страницы и представительские тексты на русском языке: лендинги, промостраницы, страницы продукта, услуги, тексты о себе, профили, отклики, страницы компании, миссии и пресс-релизы. Use this skill whenever the user asks to write, audit, structure, or improve a Russian landing page, promo page, sales page, product page, service page, about-me text, job application, company page, mission, press release, homepage copy, or text describing services. Триггеры: «лендинг», «промостраница», «продающий текст», «страница продукта», «текст о себе», «отклик на вакансию», «о компании», «миссия», «пресс-релиз», «опиши мои услуги», «текст для главной». Для коротких постов см. skill post; для статей и лонгридов — statya; для текстов интерфейса — ux-copy.
Короткие русскоязычные посты для Telegram, LinkedIn, Instagram и других соцсетей: пост с нуля, пост из статьи, анонс в канал, переписывание первой строки, усиление хука и финала. Use this skill whenever the user asks to write, rewrite, draft, shorten, or improve a Russian social media post, Telegram channel post, LinkedIn post, Instagram caption, channel announcement, post from an article, or first-line hook. Триггеры: «пост в телеграм», «пост для канала», «пост в LinkedIn», «пост в Instagram», «короткий пост», «напиши пост», «анонс в канал», «пост из статьи», «первая строка не цепляет», «перепиши пост». Для статей и лонгридов см. skill statya; для лендингов и продающих страниц — promo; для интерфейса — ux-copy.
Builds end-to-end document processing pipelines using LandingAI ADE. Covers batch and async processing, classify-then-extract workflows for mixed document types, RAG pipelines with vector DB ingestion, database integration (Snowflake, CSV, DataFrames), visualization (bounding box overlays, cropped chunk images, word-level annotation), and Streamlit UIs. Use when composing ADE parse/extract/split operations into multi-step pipelines, processing document batches in parallel, loading extraction results into databases, or visualizing/annotating extracted content. Complements the document-extraction skill (which covers single ADE SDK operations); use when those operations need to be chained into workflows or when word-level grounding, bounding box visualization, or annotation is required.
Parses, extracts, and classifies documents using LandingAI's Agentic Document Extraction (ADE). Covers the v2 APIs powered by DPT-3 (Parse and Extract, sync and async jobs with service tiers, hierarchical block structure, Markdown ranges, normalized bounding boxes) and the v1 APIs (page classification with Classify, table of contents generation with Section, schema generation with Build Extract Schema, document splitting with Split, plus v1 Parse and Extract for Office formats and password-protected files). Use when parsing documents into structured Markdown, extracting fields with a JSON Schema or Pydantic model, classifying pages or splitting mixed document batches, generating a table of contents, processing large files asynchronously, migrating from ADE v1 to v2, or when the user mentions blocks, chunks, grounding, bounding boxes, ranges, spans, or highlighting where data appears in a document.
> Plan and control time-step policies for transient simulations — couple CFL and physics-based stability limits with adaptive stepping, ramp initial transients through sharp gradients or phase changes, schedule output intervals and checkpoint cadence, and plan restart strategies for long-running jobs. Use when choosing dt for a new simulation, diagnosing adaptive time-step oscillations, deciding checkpoint frequency to minimize lost work, or setting up output schedules aligned with physical time scales, even if the user only says "my run is too slow" or "how often should I save."
> Select and apply numerical differentiation schemes for PDE and ODE discretization — generate finite-difference stencils at arbitrary order and accuracy, choose between central, upwind, compact (Pade), and spectral methods, handle boundary stencils, and estimate truncation error scaling. Use when discretizing spatial derivatives, picking a scheme for advection- or diffusion-dominated problems, building custom stencils for nonstandard operators, or comparing dispersion and dissipation properties of candidate schemes, even if the user just says "how do I approximate this derivative" or "my solution is too diffusive."
> Select and configure nonlinear solvers for root-finding f(x)=0, optimization min F(x), and least-squares problems — choose among Newton, Newton-Krylov, quasi-Newton (BFGS, L-BFGS), Broyden, Anderson acceleration, and Levenberg-Marquardt methods, configure line search or trust-region globalization, diagnose convergence rate (quadratic, linear, stagnated), and assess Jacobian quality and conditioning. Use when a Newton solver converges slowly or diverges, choosing between line search and trust region, debugging nonlinear iteration failures in FEM or phase-field codes, or selecting a solver for large-scale unconstrained optimization, even if the user only says "my Newton iterations aren't converging."
> Plan and evaluate mesh generation for numerical simulations — estimate grid resolution from physics scales (interface width, boundary layers, wavelengths), check aspect ratios and skewness against quality thresholds, choose between structured, unstructured, and adaptive mesh refinement strategies, and compute grid sizing for 1D/2D/3D domains. Use when setting up a new mesh, diagnosing poor solver convergence caused by mesh quality, deciding how many points to place across a phase-field interface or boundary layer, or preparing a mesh convergence study, even if the user only asks "what resolution do I need" or "why is my solver failing."
> Select and configure linear solvers for Ax=b systems arising in numerical simulations — choose between direct (LU, Cholesky) and iterative (CG, GMRES, BiCGSTAB, MINRES) methods, analyze sparsity patterns and matrix conditioning, recommend preconditioners (AMG, ILU, IC), apply row/column scaling, and diagnose convergence stagnation from residual histories. Use when setting up a linear solve for FEM/FVM assembly, debugging slow or stalled Krylov iterations, choosing a preconditioner for SPD or nonsymmetric systems, or investigating ill-conditioning, even if the user only says "my solver is slow" or "GMRES won't converge."
> Analyze numerical stability for time-dependent PDE simulations — check CFL and Fourier criteria, perform von Neumann stability analysis, detect stiffness, evaluate matrix conditioning, and recommend explicit vs implicit time-stepping schemes. Use when selecting time steps, diagnosing numerical blow-up or solver divergence, checking convergence criteria, or evaluating scheme stability for advection, diffusion, or reaction problems, even if the user doesn't explicitly mention "stability" or "CFL."
> Select and configure time integration methods for ODE and PDE simulations — choose among explicit Runge-Kutta, BDF, Rosenbrock, and Adams families, set relative and absolute error tolerances, implement adaptive step-size control with P/PI step-size controllers, plan IMEX operator splitting for mixed stiff and non-stiff terms, and estimate splitting errors. Use when picking an integrator for a new simulation, diagnosing step rejections or tolerance failures, setting up operator splitting for phase-field or reaction-diffusion problems, or deciding between explicit and implicit time marching, even if the user only says "my solver keeps rejecting steps" or "which ODE method should I use."
> Perform spatial and temporal convergence analysis for solution verification — compute observed convergence orders from grid or timestep refinement studies, apply Richardson extrapolation to estimate discretization error, and calculate the Grid Convergence Index (GCI) per ASME V&V 20 standards. Use when verifying that a numerical solution converges at the expected rate, estimating the error on the finest mesh, checking whether grids are in the asymptotic range, or preparing formal verification reports, even if the user only asks "is my mesh fine enough" or "how accurate is my solution."
> Create FAIR-minded reproducibility bundles for materials simulations by collecting input/output file inventories, hashes, units, engine versions, structure identifiers, provenance, and NOMAD/OPTIMADE/Materials Project friendly metadata. Use before publishing, sharing, archiving, or handing simulation results to another agent.
> Diagnose HPC runtime and scheduler problems for materials simulations, including MPI/OpenMP/GPU layout, modules, CUDA/Kokkos hints, scratch paths, walltime, job arrays, restart strategy, scheduler portability, and resource mismatch. Use when an ALREADY-SUBMITTED job failed, ran slowly, got killed, or behaved differently on a cluster than on a workstation. To author or vet a NEW submission script before running, use slurm-job-script-generator instead.
> Parse, navigate, and query materials science ontology structures — browse class hierarchies, inspect individual classes and their properties, look up object and data property definitions with domain/range, search for ontology terms by keyword, and parse or summarize raw OWL/XML files. Currently supports CMSO and ASMO; the broader OCDO ecosystem (CDCO, PODO, PLDO, LDO) is planned. Use when exploring what classes or properties an ontology provides, finding the right CMSO term for a crystal structure or simulation concept, understanding parent-child class relationships, or onboarding to an unfamiliar materials ontology, even if the user only says "what ontology terms describe my FCC copper simulation" or "show me the CMSO class hierarchy."
> Generate correct, copy-pasteable SLURM sbatch job scripts and sanity-check HPC resource requests — configure nodes, MPI tasks, OpenMP threads, memory (per-node or per-cpu), GPUs, walltime, partitions, modules, and environment variables, with automatic detection of conflicting directives and oversubscription. Use when preparing a SLURM submission script, deciding between pure MPI and hybrid MPI+OpenMP layouts, standardizing #SBATCH directives across a team, debugging why a job won't launch or gets killed, or setting up GPU-accelerated simulation jobs, even if the user only says "I need to run this on the cluster" or "my job keeps getting killed."
> Map materials science terms, crystal structures, and sample descriptions to standardized ontology classes and properties — resolve natural-language concepts to ontology entries with confidence scores, translate Bravais lattice types, space groups, and lattice constants into ontology-compliant annotations, and produce full sample metadata from structured descriptions. Supports any ontology in ontology_registry.json (CMSO, ASMO, etc.). Use when annotating simulation inputs with FAIR metadata, translating "BCC iron" or "FCC copper" into formal ontology terms, preparing machine- readable sample descriptions, or bridging between lab vocabulary and ontology vocabulary, even if the user only says "what CMSO terms describe my material" or "annotate this sample for me."
> Rigorously evaluate an Agent Skill end-to-end across ANY coding-agent CLI — verify its scripts emit the documented numbers (deterministic checks), test whether its description triggers on the right prompts, and measure whether an agent following the SKILL.md beats a no-skill baseline (with/without pass-rate delta, mean ± stddev, benchmarked). Use whenever you need to test, benchmark, validate, grade, or quantify a skill's quality, check if a skill "actually works," compare two skill versions, optimize a skill's triggering, or set up an eval suite — even if the user just says "is this skill any good," "does my skill work," or "benchmark this skill." Drives Claude Code, OpenAI Codex, Antigravity (agy), Cursor, GitHub Copilot, Amp, opencode, or Grok in headless mode.
> Validate material sample annotations against ontology constraints — check that class names and property names exist in the ontology, verify domain and range consistency for object property relationships, assess annotation completeness (required, recommended, and optional properties), and flag unknown or misspelled terms. Use when verifying that CMSO or other ontology annotations are correct before publishing, checking whether all required properties are present for a class like Crystal Structure or Unit Cell, auditing relationship triples between instances, or catching annotation errors early in a FAIR data workflow, even if the user only says "is my annotation correct" or "what am I missing."
> First-response triage for cross-code (LAMMPS/VASP/QE/MOOSE) simulation exploding energy, unstable timestep, pressure blow-up, missing potentials/pseudopotentials, memory exhaustion, process crash/segfault, corrupted output, incomplete runs) and proposes a SAFE one-change-at-a-time RETRY LADDER with explicit STOP CONDITIONS, prioritizing evidence preservation. Use as the immediate first response to a failed or suspicious run. For deep per-stage validation, conservation/physical-bounds checks, and "can I trust these results" analysis, defer to the simulation-validator skill.
> Plan molecular dynamics post-processing for materials simulations, including RDF, MSD and diffusion, VACF/VDOS, coordination numbers, bond-angle distributions, stress-strain curves, equilibration detection, PBC unwrapping, and trajectory format choices. Use before writing MD analysis scripts or trusting trajectory-derived results.
> Identify computational bottlenecks, analyze parallel scaling, estimate memory requirements, and generate optimization recommendations for materials simulations — parse timing logs to find dominant phases (solver, assembly, I/O), evaluate strong and weak scaling efficiency, profile memory from mesh and field parameters, and detect bottlenecks with actionable fix suggestions. Use when a simulation is running slower than expected, investigating MPI scaling efficiency, planning HPC resource allocation, deciding whether to tune the preconditioner or reduce I/O frequency, or estimating if a problem fits in available RAM, even if the user only says "my simulation is too slow" or "how many nodes do I need."
> Explore and optimize simulation parameters via design of experiments (DOE), sensitivity analysis, and optimizer selection — generate Latin Hypercube, quasi-random, or factorial sample plans, rank parameter influence with sensitivity scores, recommend Bayesian optimization, CMA-ES, or gradient- based methods based on dimension and budget, and fit surrogate models for expensive evaluations. Use when calibrating material properties against experimental data, planning a parameter sweep, performing uncertainty quantification, or choosing an optimization strategy for a simulation with a limited evaluation budget, even if the user only says "which parameters matter most" or "how do I calibrate my model."
> Orchestrate multi-simulation campaigns — generate parameter sweep configurations (grid, linspace, or Latin Hypercube sampling), initialize and track batch job campaigns, monitor job completion status, and aggregate results with summary statistics across all runs. Use when running a parameter study across dt, kappa, or other simulation inputs, managing dozens or hundreds of simulation configurations, combining outputs from completed batch runs to find the best result, or automating the generate-run-collect workflow for systematic studies, even if the user only says "I need to try many parameter combinations" or "how do I organize a sweep."
> Extract, analyze, and summarize simulation output data — pull spatial fields at specific timesteps, compute time-series trends and detect steady state, extract line profiles through the domain, generate statistical summaries and distributions, calculate derived quantities (gradients, fluxes, volume fractions, interface area), compare results against analytical solutions or experimental data, and produce automated analysis reports. Use when interpreting finished simulation results, checking mass or energy conservation, comparing two runs or meshes, extracting interface profiles from phase-field output, or preparing publication-quality analysis, even if the user only says "what do my results look like" or "did my simulation reach steady state."
> Validate simulations across three stages — run pre-flight checks on configuration files (parameter ranges, required fields, disk space), monitor runtime logs for residual growth, NaN/Inf, and adaptive dt collapse, and perform post-flight validation of results (physical bounds, mass/energy conservation, convergence). Diagnose failed simulations with probable-cause analysis and recommended fixes. Use when preparing to launch a simulation, checking whether a running job is healthy, verifying that finished results are trustworthy, or debugging a crash or blow-up, even if the user only says "my simulation crashed" or "can I trust these results."
> Plan verification and validation campaigns for simulation codes using manufactured solutions, canonical benchmark problems, grid/time refinement, uncertainty propagation, and pass/fail acceptance criteria. Use when an agent needs to prove a solver, model, or result is trustworthy rather than only plausible.
> Map computational materials tasks onto workflow engines such as atomate2, jobflow, AiiDA, pyiron, or a simple one-off script. Use when deciding how to structure a reproducible campaign, DAG, restart strategy, provenance record, storage layout, or migration path from ad hoc scripts to managed workflows.
>- Use this skill to run the genesis maintainer-side eval suite against genesis PR, when changing the genesis catalogue (architectural-patterns, primitives, design-patterns, refactor-patterns, composition-substrate, pattern-tradeoffs, SKILL.md), or when the operator asks to "run evals" or "regenerate the eval matrix". This skill orchestrates parallel cold sub-agent spawns via the harness's task tool, scores deterministically, and converges P>=0.8 / N>=0.8 / R==1.0 within max 3 iteration loops. This skill is contributor-only -- it lives under dev/skills/ (OUTSIDE .apm/) and is NOT shipped inside the user-facing skills/genesis/ bundle (BUNDLE LEAKAGE discipline). See "Why this lives outside .apm/" below.
Claude Skills — questions
Answers built from the skills we actually parsed.