WT
Community@skyllwt · China
Peking University major in engineering.
Agent Skills by WT
Showing 53 vetted skills indexed across 2 GitHub repositories.
reset
Resets wiki state to a clean scaffold by selectable scope with confirmation.
edit
Adds or removes raw sources and updates wiki content per user request.
ideate
Generates research ideas through a five-phase pipeline grounded in a wiki knowledge base.
research
Orchestrates end-to-end research pipelines from idea discovery through experiments to paper writing.
daily-arxiv
Generates daily arXiv paper recommendations and digests from a research wiki profile.
init
Builds a research wiki from local papers and discovered literature via parallel ingest.
exp-design
Designs iterative experiment suites with ablations, sensitivity analysis, and baselines for research ideas.
paper-draft
Drafts LaTeX paper sections from a wiki-based paper plan with figures and verified BibTeX.
discover
Generates ranked paper shortlists from anchor papers, topics, wiki state, or venue lists.
ask
Answers questions over a research wiki by retrieving pages and synthesizing cited responses.
exp-pilot-run
Executes pilot experiments from Pilot Spec YAML files on local or remote GPUs.
exp-status
Monitors running experiments via screen sessions and SSH, reporting status and advancing pipelines.
paper-compile
Compile LaTeX papers to PDF with latexmk, auto-fix errors, and verify submission requirements.
novelty
Validates research idea novelty via multi-source search and cross-model review scoring.
review
Reviews research artifacts with an independent LLM and outputs structured scores and improvement suggestions.
paper-plan
Compiles paper outlines from a research wiki idea graph with evidence maps and citation plans.
refine
Iteratively improves research artifacts through repeated review, fix, and wiki update cycles.
exp-eval
Evaluates completed experiments against linked idea hypotheses and updates wiki knowledge graph.
survey
Generates Related Work sections from wiki knowledge with grouped citations and LaTeX output.
exp-run
Prepares, deploys, monitors, and collects results from planned ML experiments.
check
Scan wiki pages and graph files to detect structural issues and generate prioritized lint reports.
exp-pilot-eval
Evaluates pilot experiment results and updates idea pages with pass, fail, or inconclusive verdicts.
poster
Generates self-contained HTML academic posters from LaTeX paper sources.
rebuttal
Generates evidence-traced academic rebuttals by mapping reviewer concerns to wiki ideas and experiments.