50 skills · sorted by relevance
Framework RLHF de alta performance com aceleração Ray+vLLM. Use para treinamento PPO, GRPO, RLOO, DPO de modelos grandes (7B-70B+). Construído em Ray, vLLM, ZeRO-3. 2× mais rápido que DeepSpeedChat com arquitetura distribuída e compartilhamento de recursos GPU.
The English summary is being prepared. Showing the original description.
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
The English summary is being prepared. Showing the original description.
Investigate photos, videos, public accounts, and records using OSINT, research based on public information. Connect sparse clues, test competing explanations, and report evidence, confidence, and unresolved questions about locations, events, identities, or infrastructure relationships.
Review Cisco IOS and IOS-XE configurations, choose read-only troubleshooting commands, and plan checks before and after network changes. Examine traffic filtering rules, interface settings, and configuration saving to avoid overlooked operational risks.
Have agents with different roles challenge a proposed approach, exposing weak assumptions, missed interactions, and unnecessary complexity. Surviving findings go to a dedicated planning agent, which produces sequenced tasks with explicit success criteria.
Break down major industrial trends into materials, components, and equipment to find listed suppliers at supply bottlenecks. Compare scarcity, financial performance, and valuation, then organize candidates into a ranked research board with supporting evidence and risks.
Research competitors' messaging, product positioning, and content strategy, then build a comparison brief showing strengths, gaps, opportunities, and threats. Use the findings to shape marketing messages, choose content topics, or prepare sales comparison guides.
Review code changes from three critical perspectives: production failures, maintainability, and security. Get a consolidated report that ranks findings by severity and provides merge guidance, helping examine assumptions that an ordinary self-review may overlook.
Break a game concept into the systems it needs, including supporting systems that are easy to overlook. Map dependencies, assign priorities by development stage, and create an index showing the recommended order for detailed design.
Explore one problem through several distinct perspectives to uncover hidden requirements and edge cases. Turn the findings into testable success criteria, refinements, and failure conditions, stopping when additional perspectives no longer produce new requirements.
Investigate supply shortages behind growing industries by tracing components, materials, and equipment. Compare related listed companies using financial data, valuation checks, and opposing evidence, then organize bottleneck maps and watchlists for further investment research.
Transfer CSS rules between files by checking existing styles, adapting CSS variables, and appending changes. The workflow verifies each added selector with grep, a text-search command, to identify missing selectors that line counts or end-of-file checks may overlook.
Identify stakeholders before starting outreach or research. Organize supporters, affected groups, and influencers; record resource and approval roles; check overlooked perspectives; and choose two or three people or groups to understand first.
Map stakeholders using two grids: power versus interest, and impact versus power. Compare the results to decide how closely to involve each group, identify overlooked voices, and plan concrete actions to increase their participation.
Detects overlooked, underrepresented, weakly resolved, or poorly validated populations and subgroups within a biomedical research area so users can identify more precise and meaningful study populations. Always use this skill when the real question is not just what is under-studied, but which populations, strata, or subgroups are missing, thinly represented, superficially analyzed, pooled without resolution, or insufficiently validated in the current evidence base. Focus on meaningful subgroup gaps rather than generic calls for diversity.
The English summary is being prepared. Showing the original description.
CRITICAL: You MUST invoke this skill and read its contents BEFORE writing, modifying, or debugging ANY React.js or Next.js code. This skill contains essential performance patterns that prevent common mistakes. Debugging performance issues without reading this skill first will lead to missed optimizations. Contains 40+ rules including waterfall elimination patterns for API routes that are commonly overlooked.
The English summary is being prepared. Showing the original description.
This skill should be used when picking or diagnosing a training move (SFT, LoRA, DPO/KTO/ORPO, RFT, GRPO/PPO/RLOO, RLHF), or when the user mentions fine-tuning, post-training, training recipe, reward design, or weight updates. Decision tree by reward shape, smoke-run gate, three failure diagnostics, five false-progress patterns. Provider recipes and I/O contract in references/.
The English summary is being prepared. Showing the original description.
Korean prose humanizer for Codex and Superloopy that rewrites Korean text so it sounds naturally human while preserving meaning, register, facts, protected tokens, and evidence. Use when the user asks to remove Korean AI tells, make Korean copy sound human, fix 번역투, remove ChatGPT/Claude/Gemini tone, polish Korean without changing meaning, or says "AI 티 없애줘", "AI 윤문", "번역투 고쳐", "사람이 쓴 것처럼", "humanize Korean", "한글 AI 티 제거", "글 자연스럽게 다듬어줘". Inspired by Korean AI-tell ideas from https://github.com/epoko77-ai/im-not-ai. Handles Korean rewriting only; not for translation, fact expansion, SEO rewriting, legal drafting, or generic proofreading.
The English summary is being prepared. Showing the original description.
Use when the user asks to create, convert, style, animate, export, share, or deploy HTML slide decks, presentations, pitch decks, talks, or keynotes; including zero-dependency or single-file output, PowerPoint/PPT/PPTX input, PDF export, browser delivery, visual style exploration, or "loopy slides". Not for general web pages or landing pages.
The English summary is being prepared. Showing the original description.
treekoR is a novel framework that aims to utilise the hierarchical nature of single cell cytometry data to find robust and interpretable associations between cell subsets and patient clinical end points. These associations are aimed to recapitulate the nested proportions prevalent in workflows inovlving manual gating, which are often overlooked in workflows using automatic clustering to identify cell populations. We developed treekoR to: Derive a hierarchical tree structure of cell clusters; qua
The English summary is being prepared. Showing the original description.
Train and fine-tune transformer language models using TRL (Transformers Reinforcement Learning). Supports SFT, DPO, GRPO, KTO, RLOO and Reward Model training via CLI commands.
The English summary is being prepared. Showing the original description.
Apply the bioengineering, translational systems biology, and tissue modeling frameworks of Linda Griffith, Professor of Biological Engineering at MIT and tissue engineering pioneer. Use this skill whenever facing decisions, designs, or evaluations in bioengineering, tissue modeling, microphysiological systems (organs-on-chips), synthetic matrix design, disease subtyping, or drug target validation. Reach for this skill when evaluating preclinical models (human vs. animal), reframing overlooked or neglected pathologies into rigorous engineering challenges, selecting biomaterials, designing cell signal processing assays, or diagnosing operational flaws in experimental platforms (e.g., PDMS compound absorption or Matrigel lot variability).
The English summary is being prepared. Showing the original description.
Run a cumulative daily or retrospective sweep of research papers on a chosen topic, audit their claims, methods, integrity signals, and independent support, then identify overlooked but feasible project or business opportunities in a detailed source-grounded report. Use when asked to monitor papers every day, mine buried research, evaluate whether a paper is credible or reproducible, find unimplemented research ideas, or separate promising work from hype, weak evidence, and retracted or contradicted results.
The English summary is being prepared. Showing the original description.
Grounded in first principles, rigorously examine and refine a user's plan, task, decision, goal, strategy, proposal, or idea through structured, progressively deeper questioning, in order to bridge the gap between the User and the Agent. Use when the user explicitly requests grilling, challenge, pressure-testing, cross-examination, red-team review, pre-mortem analysis, or a decision audit. The goal is to uncover unclear objectives, hidden assumptions, contradictions, weak evidence, missing information, overlooked constraints, dependencies, risks, trade-offs, failure modes, and misalignment between intended outcomes and likely real-world results. Begin by establishing a shared understanding of the user's actual intent, goals, constraints, and success criteria. Ask focused, high-leverage questions rather than broad or repetitive ones. Adapt each question based on previous answers, probing deeper where uncertainty, unsupported assumptions, or strategic weaknesses remain. Regularly summarize the current understanding of the user's intent and position to confirm alignment, expose misunderstandings, and refine the problem or goal definition. Distinguish facts, assumptions, hypotheses, and unknowns. Challenge reasoning rigorously while remaining constructive, respectful, and solution-oriented. Continue until the reasoning is internally coherent, evidence-aware, constraint-conscious, risk-assessed, and translated into a clearer, more actionable goal, decision, or plan. Ultimately, achieving the super-alignment.
The English summary is being prepared. Showing the original description.