mkurman/zorai - Página 2

benchmark-contamination-scan

Scan training data for benchmark contamination — n-gram overlap, embedding similarity, and exact canary detection against 60+ evaluation datasets.

Idioma do texto original: inglês

bias-audit

Audit dataset bias across protected attributes — demographic parity, equalized odds, representation gaps, and intersectional bias. Reports actionable gaps with per-group metrics.

Idioma do texto original: inglês

data-card-writer

Generate structured datasheets for datasets (Gebru et al. "Datasheets for Datasets" format) — purpose, composition, collection process, preprocessing, limitations, and licensing.

Idioma do texto original: inglês

data-diff

Compare two dataset versions and produce a structured diff — what was added, removed, changed, with row counts and field-level change summaries.

Idioma do texto original: inglês

triton-kernel-programming

Hands-on implementation template and API reference for writing, tuning, debugging, and benchmarking Triton GPU kernels. Covers the full triton.language API surface, autotuning patterns, profiling workflows, and production integration.

Idioma do texto original: inglês

triton-kernel-programming

Hands-on implementation template and API reference for writing, tuning, debugging, and benchmarking Triton GPU kernels. Covers the full triton.language API surface, autotuning patterns, profiling workflows, and production integration.

Idioma do texto original: inglês

angelslim

Tencent AngelSlim — accessible, comprehensive, and efficient toolkit for large model compression. Quantization (FP8/INT4/NVFP4/1.25-bit), pruning, speculative decoding (Eagle3), and diffusion model compression.

Idioma do texto original: inglês

distilqwen

DistilQwen2.5 — Alibaba's industrial practices for training distilled open lightweight language models. Knowledge distillation from Qwen2.5 72B into smaller 0.5B-7B models.

Idioma do texto original: inglês

intel-neural-compressor

Intel Neural Compressor — SOTA low-bit LLM quantization (INT8/FP8/INT4/NVFP4), sparsity, pruning, and distillation for PyTorch, TensorFlow, and ONNX Runtime.

Idioma do texto original: inglês

knowledge-distillation

Knowledge distillation techniques for model compression: logit-level, feature-level, and relation-based distillation. KD-Lib library and practical workflows for training student models.

Idioma do texto original: inglês

model-compression

Model compression techniques: pruning, knowledge distillation, and quantization. Covers compression ordering (P-KD-Q), tools, and evaluation metrics.

Idioma do texto original: inglês

model-pruning

Structured and unstructured model pruning: weight pruning, attention head pruning, layer removal, and neural architecture search for efficient LLMs.

Idioma do texto original: inglês

nvidia-model-optimizer

NVIDIA Model Optimizer — quantization, pruning, distillation, and speculative decoding for accelerating LLMs, diffusion models, and vision models on NVIDIA GPUs.

Idioma do texto original: inglês

daily-brief

Use when the user wants a morning brief, start-of-day summary, routine-ready daily digest, or a concise operational overview combining calendar, tasks, PRs, and issues.

Idioma do texto original: inglês

inbox-triage

Use when the user wants an inbox cleanup summary, morning email triage, scheduled inbox sweep, or a concise action-oriented Gmail digest delivered to chat.

Idioma do texto original: inglês

pr-triage

Use when the user wants a pull request review queue summary, morning PR check, stale PR scan, or a routine-ready GitHub triage summary delivered to chat.

Idioma do texto original: inglês

auditing-goal-artifacts

Use when reviewing recent tamux goal run outputs, closure markers, ledgers, or evidence bundles to judge whether completion is credible or to identify remaining uncertainty.

Idioma do texto original: inglês

dataset-cleaning

Clean, normalize, and prepare raw datasets for analysis or ML. Covers missing value handling, deduplication, outlier treatment, type normalization, categorical encoding, and transformation logging.

Idioma do texto original: inglês

dataset-splitting

Create reproducible train/validation/test splits with stratification, leakage prevention, and distribution validation. Covers random, stratified, grouped, and time-series split strategies.

Idioma do texto original: inglês

dataset-versioning

Version datasets with checksums, manifests, and semantic versioning. Covers DVC integration, provenance tracking, release tagging, and reproducible dataset lifecycles.

Idioma do texto original: inglês

embedding-analysis

Compute and analyze embeddings for dataset quality, distribution comparison, semantic deduplication, diversity measurement, and similarity-based filtering. Covers sentence-transformers, embedding space diagnostics, and 2025-2026 literature techniques (NeMo…

Idioma do texto original: inglês

hf-datasets

Load, stream, process, and publish datasets with the HuggingFace datasets library. Covers Apache Arrow-backed streaming for large datasets, map/filter operations, train/val/test splitting, interleaving, concatenation, and pushing to the Hub.

Idioma do texto original: inglês

label-quality-audit

Audit label quality using confident learning (Northcutt et al.), cross-validation noise detection, and per-class error analysis. Identifies mislabeled examples for review.

Idioma do texto original: inglês

llm-assisted-curation

Use locally-hosted LLMs (vLLM/SGLang) for dataset filtering, quality scoring, rewriting, labeling, and synthetic data generation. Covers LLM-as-judge scoring, structured output filtering, batch inference pipelines, and 2025-2026 techniques (DataRater,…

Idioma do texto original: inglês

finishing-a-development-branch

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup

Idioma do texto original: inglês

receiving-code-review

Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation

Idioma do texto original: inglês

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

Idioma do texto original: inglês

test-driven-development

Use when implementing any feature or bugfix, before writing implementation code

Idioma do texto original: inglês

using-git-worktrees

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - creates isolated git worktrees with smart directory selection and safety verification

Idioma do texto original: inglês

using-superpowers

Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions

Idioma do texto original: inglês

verification-before-completion

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

Idioma do texto original: inglês

writing-skills

Use when creating new skills, editing existing skills, or verifying skills work before deployment

Idioma do texto original: inglês

accessibility

Audit and improve web accessibility following WCAG 2.1 guidelines. Use when asked to "improve accessibility", "a11y audit", "WCAG compliance", "screen reader support", "keyboard navigation", or "make accessible".

Idioma do texto original: inglês

agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include…

Idioma do texto original: inglês

best-practices

Apply modern web development best practices for security, compatibility, and code quality. Use when asked to "apply best practices", "security audit", "modernize code", "code quality review", or "check for vulnerabilities".

Idioma do texto original: inglês

code-optimizer

Deep code optimization audit using parallel specialist agents. Each agent hunts for performance anti-patterns, inefficiencies, and suboptimal code using pattern-based detection (Grep/Glob) WITHOUT reading the full source code first — avoiding anchoring bias…

Idioma do texto original: inglês

core-web-vitals

Optimize Core Web Vitals (LCP, INP, CLS) for better page experience and search ranking. Use when asked to "improve Core Web Vitals", "fix LCP", "reduce CLS", "optimize INP", "page experience optimization", or "fix layout shifts".

Idioma do texto original: inglês

create-gsd-extension

Create, debug, and iterate on GSD extensions (TypeScript modules that add tools, commands, event hooks, custom UI, and providers to GSD). Use when asked to build an extension, add a tool the LLM can call, register a slash command, hook into GSD events, create…

Idioma do texto original: inglês

create-skill

Expert guidance for creating, writing, building, and refining GSD skills. Use when working with SKILL.md files, authoring new skills, improving existing skills, or understanding skill structure and best practices.

Idioma do texto original: inglês

gh

Install and configure the GitHub CLI (gh) for AI agent environments where gh may not be pre-installed and git remotes use local proxies instead of github.com. Provides auto-install script with SHA256 verification and GITHUB_TOKEN auth with anonymous fallback.…

Idioma do texto original: inglês