VeryMath

Route AI4Math computational mathematics tasks to the appropriate bundled skill package.

Use when reproducing, deploying, diagnosing, tuning, visualizing, or reporting computational math research code with a coding agent

This skill should be used when the user asks to search for external algorithms, find related implementations, discover alternative approaches, or look up papers and code for computational math optimization methods.

This skill should be used when the user asks to tune parameters, run hyperparameter search, optimize algorithm settings, or find the best configuration for a computational math method.

Use when classifying a computational math repository or task into a broad domain before choosing specialist Skills, references, runtimes, and validation evidence.

Use when a user starts an end-to-end computational math research-code reproduction, deployment, tuning, visualization, or reporting workflow with a coding agent.

This skill should be used when the user asks to detect what algorithm family a piece of code uses, identify optimization methods, or classify computational math approaches such as ADMM, PPA, proximal gradient, or primal-dual methods.

This skill should be used when the user asks to set up an environment, check dependencies, prepare deployment, or analyze how to run code in a specific computational environment.

Route AI4Math paper-reading tasks to the appropriate bundled skill package.

Use when extracting reusable AI4Math research skills from papers, especially PDF-to-SkillCard workflows, paper triage, PDF preparation, proof-pattern extraction, and cross-paper SkillCard synthesis for a coding-agent-neutral paper-to-skill workflow.

Cluster, deduplicate, generalize, and rank proof-pattern SkillCandidates or SkillCards extracted from multiple papers. Use when a coding agent should synthesize single-paper extractions into a domain method map, merged SkillCards, and a synthesis report for an AI4Math skill library.

Search for papers that may provide new methods, new theory, sharper bounds, different assumptions, proof techniques, lower bounds, abstractions, or cross-domain transfers for a math research direction. Use when a coding agent should look beyond nearest-neighbor related work and generate innovation_candidates.json with reasons and risks.

Download open-access research-paper PDFs selected from reading_plan.json, candidate_papers.json, or innovation_candidates.json. Use when a coding agent should save paper.pdf, metadata.json, and a bilingual download report for confirmed paper IDs without bypassing paywalls, and before PDF-to-Markdown conversion or paper-to-skill extraction.

Extract reusable mathematical research skills from Markdown papers, especially proof-pattern skills such as descent lemmas, inequality relaxations, telescoping bounds, error decompositions, Lyapunov arguments, concentration bounds, Galerkin/Cea arguments, Aubin-Nitsche duality, and Lax-Milgram well-posedness. Use when a coding agent is asked to turn a math paper, proof, theorem section, convergence analysis, stability analysis, or error analysis into SkillCandidate JSON, SkillCard YAML, or a paper-to-skill report without relying on external LLM APIs.

Run the complete AI4Math paper-to-skill workflow from a user-provided PDF or Markdown paper. Use when a coding agent should preserve the original paper, convert PDF to paper.md when needed, extract proof-pattern SkillCandidates and SkillCards with line evidence, and optionally route reviewed cards into cross-paper synthesis.

Rank candidate research papers into must_read, should_read, maybe_read, and skip groups for a paper-to-skill workflow. Use when a coding agent should combine related-paper and innovation candidates, judge relevance, novelty, proof-pattern value, reading cost, access status, and produce a reading_plan.json with user confirmation checkpoints.

Route AI4Math mathematical paper-writing tasks to the bundled paper-writing skill package.

Use when drafting, revising, or auditing source-grounded mathematical paper text, including abstracts, introductions, related work, theorem exposition, experiment narratives, revision plans, and response letters.

Use when auditing mathematical paper drafts for supported claims, missing citations, overclaims, proof status, experiment support, or source-to-text traceability.

Use when improving mathematical formulas, displayed derivations, theorem/proof environments, equation alignment, cases blocks, long displays, or formula-to-prose readability in paper drafts.

Use when checking LaTeX paper projects for compilation, latexmk logs, undefined references, citation issues, duplicate labels, macro/package hygiene, layout warnings, floats, arXiv, or venue compatibility.

Use when checking mathematical paper drafts for notation drift, variable reuse, undefined symbols, inconsistent domains, index conventions, or scalar/vector/matrix ambiguity.

Use when turning mathematical notes, theorem statements, proof sketches, experiments, or reading outputs into a paper skeleton, section plan, result dependency map, or contribution architecture before prose drafting.

Use when checking mathematical paper results for assumptions, quantifiers, domains, dependency fit, edge cases, external theorem use, proof coverage, or theorem-to-claim consistency.

Route AI4Math optimization tasks to the appropriate bundled skill package.

Models and solves linear programs (LP) with multiple solvers (COPT, Gurobi, MOSEK, CPLEX, HiGHS (scipy/highspy), CLARABEL, PuLP/CBC, CVXOPT, OR-Tools/GLOP, GLPK, SoPlex, lpsolve, ECOS). Restates natural-language word problems into LP before building and solving. When no local solver is available, searches GitHub for open-source LP solving code. Use when the user mentions linear programming, LP, word problems, applications, story problems, natural-language descriptions, 应用题, 线性规划文字题, matrix-form inequalities or equalities, and variable bounds.

Models and solves second-order cone programs and problems convertible to SOCP form using cvxpy with multiple solver backends (COPT, Gurobi, MOSEK, CPLEX, CLARABEL, ECOS, SCS, CVXOPT, COSMO, OSQP). Handles linear objectives with second-order cone constraints, as well as problems reducible to SOCP such as portfolio optimization, robust optimization, and engineering design. When no local solver is available, searches GitHub for open-source SOCP solving code.

Use when Codex needs to solve, reproduce, test, compare, generate examples for, or diagnose CDOpt manifold optimization workflows, including CDOpt package smoke tests, official problem description cards, Stiefel dictionary learning, SciPy wrappers, PyTorch/JAX/NumPy CDOpt examples, dependency checks, tiny CPU validation runs, comparison plans, or JSON result reporting.

Models and solves mixed-integer linear programs, including binary, integer, and continuous decision variables with linear objectives and constraints.

Unified solver detection, installation, selection, and license configuration for OR optimization skills (LP, MIP, SOCP). One-stop solver management: detect installed packages, install missing solvers with proper priority, select the best available solver by problem type, and troubleshoot common errors. Covers 20 solvers across commercial (COPT, Gurobi, MOSEK, CPLEX) and open-source (HiGHS, SCIP, CLARABEL, CBC, ECOS, SCS, etc.). Use when the LP, MIP, or SOCP skill needs to set up a solver environment, or when the user asks about solver installation, solver comparison, license setup, or solver selection.

Models and solves linear programs (LP) with Cardinal Optimizer (coptpy). Restates natural-language word problems into LP before building and solving. Use when the user mentions COPT/coptpy, linear programming, LP, word problems, applications, story problems, natural-language descriptions, 应用题, 线性规划文字题, or matrix-form constraints such as inequality systems, equality systems, and variable bounds.

Route AI4Math automated mathematical research tasks to normalized skill packages in this repository.

Use when a coding agent should clone, patch, set up, and supervise FrenzyMath Rethlas on a machine that already has OpenCode installed and configured.

Use when Codex needs to interactively deploy, configure, verify, or launch AI4Math Auto-Research with Agent Laboratory, including API-key handling, user research-topic intake, full local validation, and human review gates.

Use when Codex should act as a coding-agent-native mathematical problem discovery engine: turn fuzzy mathematical background, scattered notes, domain intuition, failed proof attempts, or immature theorem ideas into a ranked problem menu, conjecture lattice, evidence ledger, counterexample pressure, proof obligations, work orders, and a resumable research_state_packet. Use before theorem proving or formal verification when the user does not yet know which mathematical problem should be pursued.

Use when Codex should work as a coding-agent-native proof planning and review workflow: transform a candidate theorem, proof sketch, problem artifact, or proof_obligations into a proof blueprint, verifier-style report, repair hints, proof-obligation ledger patches, and a strict proof acceptance decision without treating API access as the default path.

Use when a coding agent needs Lean 4 formalization, proof repair, theorem transcription, sorry completion, Lean patch review, optional adapter-first Lean-specialist backend work, or local Lean validation.

Use for interactive Lean 4 formal verification by coding agents with reusable Lean/mathlib workspaces, theorem formalization, proof repair, sorry completion, patch review, optional adapter-first Lean-specialist backend work, and minimal failure handoff.

Use when a coding agent needs to install or verify Lean 4, elan, lake, create or reuse a mathlib workspace, configure a shared Lean environment, or run Lean readiness checks before formalization.

Route AI4Math evolving-agent and iterative skill-improvement tasks to normalized skill packages in this repository.

Use when a coding agent is asked to pursue an interactive AI4Math-Evolving/OpenEvolve goal involving initial_program.py, evaluator.py, config.yaml, metrics, logs, checkpoints, best-program artifacts, or iterative user feedback.

Use when conducting a coding-agent-driven mathematical research project that needs durable workspace state, approved goals, parallel workstreams, reviewer gates, provenance, uncertainty tracking, failed exploration records, and a final working paper.