Rajesh Kumar
Community@raja21068 · China
Currently, I am working in the Lab of Cyber Security, School of Computer Science and Engineering, UESTC
Agent Skills by Rajesh Kumar
Showing 84 vetted skills indexed across 1 GitHub repositories.
paper-writing-bench
Generate PaperWritingBench-style idea and experimental log files from AI research papers.
content-refinement-agent
Iteratively simulate peer review and revise LaTeX paper drafts with accept/revert control.
plotting-agent
Render figure_id entries from outline.json into 300-DPI PNGs with captions.
agent-research-aggregator
Aggregate agent cache logs into PaperOrchestra-ready idea.md and experimental_log.md files.
section-writing-agent
Generate missing LaTeX paper sections with verified citations and booktabs tables.
literature-review-agent
Discover and verify candidate papers via Semantic Scholar, then draft LaTeX Introduction and Related Work sections.
paper-orchestra
Generate a submission-ready LaTeX manuscript and compiled PDF from unstructured research materials via a five-agent pipeline.
outline-agent
Generate a strict JSON paper outline from methodology, logs, LaTeX templates, and venue guidelines.
paper-autoraters
Score paper drafts with citation F1 and literature-review quality JSON outputs.
proof-checker
Verify and repair LaTeX mathematical proofs with audit reports.
paper-talk
Generate a conference talk package with slides, notes, and script from a compiled paper.
paper-write
Generate a venue-formatted LaTeX paper from a plan and narrative report.
paper-poster
Generate a print-ready conference poster from a compiled LaTeX paper.
writing-systems-papers
Generate venue-aware paragraph structures for 10–12 page systems papers.
qzcli
Manage GPU compute job lifecycle on the Qizhi platform via CLI.
specification-writing
Generate jurisdiction-formatted patent specification markdown from claims and invention disclosure.
proof-writer
Produce rigorous proof packages for ML/AI theory claims with dependency maps.
paper-claim-audit
Verify quantitative paper claims against raw result files.
resubmit-pipeline
Orchestrate text-only paper resubmissions to new venues with prior reviewer reports.
serverless-modal
Run GPU training and inference workloads on Modal with scale-to-zero.
formula-derivation
Structure research formulas into a coherent derivation package with labeled steps.
training-check
Monitor WandB training metrics to detect NaN, divergence, or degraded evaluation.
meta-optimize
Analyze ARIS usage logs to propose minimal harness optimizations.
patent-novelty-check
Compare extracted claim elements against prior art references for patent novelty.