Agent Skills by Fang Jiukai
Showing 69 vetted skills indexed across 1 GitHub repositories.
proof-checker
Verify mathematical proofs in LaTeX documents and generate audit reports.
paper-write
Generate LaTeX paper sections from PAPER_PLAN.md outlines for ML and IEEE venues.
paper-poster
Converts a compiled IMRAD paper into a print-ready conference poster with LaTeX and PNG outputs.
writing-systems-papers
Generate paragraph-level blueprints for 12-page systems papers.
qzcli
Manage GPU compute jobs on the Qizhi platform via the qzcli CLI.
specification-writing
Generate a complete patent specification from claims and invention disclosure.
proof-writer
Generate structured proof packages with dependency maps for ML/AI theorems.
paper-claim-audit
Verify numerical claims in research papers against raw evidence.
serverless-modal
Automate on-demand GPU workloads on Modal serverless without SSH or Docker setup.
formula-derivation
Construct a structured derivation package from scattered notes and assumptions.
training-check
Monitor WandB training metrics periodically and alert on anomalies.
meta-optimize
Analyze ARIS usage logs and propose minimal diffs for SKILL.md optimizations.
patent-novelty-check
Analyze invention novelty and non-obviousness against public prior art.
vast-gpu
Automate vast.ai GPU provisioning, setup, execution, and teardown for ML tasks.
result-to-claim
Evaluate experiment results against research claims and route next steps.
figure-spec
Generates editable SVG architecture and workflow diagrams from FigureSpec JSON via a local renderer.
mermaid-diagram
Generate and verify Mermaid diagrams from user requirements.
semantic-scholar
Retrieve published venue papers with citation counts, venue type, and TLDR summaries.
research-wiki
Consolidate papers, ideas, experiments, and claims into a navigable research wiki.
system-profile
Profile Python scripts, processes, functions, or systems for CPU, memory, GPU, and interconnect usage.
research-refine-pipeline
Chain research-refine and experiment-plan to produce a proposal and experiment roadmap.
jurisdiction-format
Format patent documents into CN, US, and EP jurisdiction-ready filings.
experiment-audit
Audit experiment integrity across evaluation scripts, results, and paper claims.
auto-review-loop-llm
Automate iterative review, fix, and re-evaluation cycles using OpenAI-compatible LLM APIs.