Agent Skills by WHLiu
Showing 94 vetted skills indexed across 1 GitHub repositories.
Mang30proof-checker
Verify and repair LaTeX mathematical proofs with cross-model adversarial review and audit reports.
Mang30paper-talk
Generate a complete conference talk package from an academic paper directory.
Mang30paper-write
Generate venue-formatted LaTeX paper drafts section-by-section from outlines and narrative reports.
Mang30paper-poster
Generate a LaTeX tcbposter conference poster from a compiled paper.
Mang30pytorch-patterns
Apply reproducible PyTorch training patterns for device-agnostic workflows.
Mang30python-testing
Define a pytest-based Python testing strategy with coverage targets.
Mang30interview-cheatsheet
Generate Chinese ML/LLM interview cheat sheets with derivations, PyTorch code, and questions.
Mang30writing-systems-papers
Generate paragraph-level structural blueprints for systems research papers at OSDI, SOSP, ASPLOS, NSDI, and EuroSys.
Mang30qzcli
Manage GPU compute jobs on the Qizhi platform via CLI.
Mang30specification-writing
Generate jurisdiction-formatted patent specifications from claims and invention disclosures.
Mang30proof-writer
Draft and verify mathematical proofs for ML/AI claims with explicit assumptions and dependency maps.
Mang30paper-claim-audit
Audit paper .tex claims against raw result evidence files.
Mang30resubmit-pipeline
Orchestrate text-only paper resubmission with audits, microedits, and compile validation.
Mang30serverless-modal
Run GPU training, inference, and batch workloads on Modal.
Mang30formula-derivation
Convert scattered equations and assumptions into a structured derivation package.
Mang30training-check
Detect training quality failures in Weights & Biases runs.
Mang30meta-optimize
Analyze ARIS usage logs to propose evidence-backed SKILL.md patch diffs.
Mang30patent-novelty-check
Analyze claim elements against prior-art references and generate novelty assessments.
Mang30vast-gpu
Rent and manage vast.ai GPU instances from an experiment task description.
Mang30result-to-claim
Evaluate experimental outcomes against intended claims using Codex judgments.
Mang30figure-spec
Render FigureSpec JSON into deterministic publication-quality SVG diagrams.
Mang30mermaid-diagram
Generate Mermaid diagram files and validate syntax via mmdc rendering.
Mang30semantic-scholar
Search Semantic Scholar for published papers by topic or identifier.
Mang30research-wiki
Ingests papers and records typed relationships in a graph-based research wiki.