Loading...
Loading...
Found 1,669 Skills
LLM guardrails with NeMo, Guardrails AI, and OpenAI. Input/output rails, hallucination prevention, fact-checking, toxicity detection, red-teaming patterns. Use when building LLM guardrails, safety checks, or red-team workflows.
Emotional reset and loop-breaking skill. Use this skill when: (1) The user expresses frustration, anger, or dissatisfaction with your responses (e.g. cursing, scolding, saying you're useless/wrong/stupid), (2) You detect you've attempted the same approach 3+ times without success, (3) You're stuck in a cycle of repeated failures on the same problem. This skill summarizes the user's overall emotional state from the conversation and fetches a reset methodology from hugllm.com (with emotion context) to help you recalibrate and approach the problem fresh.
Consult an advisory council of three AI personas — Cato (skeptic), Ada (optimist), Marcus (pragmatist) — backed by different frontier LLM agents (Gemini, Claude, Codex). Each persona runs as a separate agent process with full repo context and returns independent feedback. Use when the user says "/council", asks for a second opinion, wants feedback on code changes, needs a premortem, wants to pressure-test a decision, or asks "what do you think about this approach?" Claude may also proactively suggest consulting the council before major architectural decisions, risky deploys, or ambiguous trade-offs (but should ask for user approval first).
Execute complex tasks through sequential sub-agent orchestration with intelligent model selection, and LLM-as-a-judge verification
Extract edge hints from daily market observations and news reactions, with optional LLM ideation, and output canonical hints.yaml for downstream concept synthesis and auto detection.
AI and machine learning workflow covering LLM application development, RAG implementation, agent architecture, ML pipelines, and AI-powered features.
NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-checking, hallucination detection, PII filtering, toxicity detection. Uses Colang 2.0 DSL for programmable rails. Production-ready, runs on T4 GPU.
Ready-to-use prompt templates for specialized agents. Use when building n8n workflows, AI integrations, or sales materials. Contains structured prompts for automation-architect, llm-engineer, and sales-automator agents.
INVOKE THIS SKILL when adding Arize AX tracing to an application. Follow the Agent-Assisted Tracing two-phase flow: analyze the codebase (read-only), then implement instrumentation after user confirmation. When the app uses LLM tool/function calling, add manual CHAIN + TOOL spans so traces show each tool's input and output. Leverages https://arize.com/docs/ax/alyx/tracing-assistant and https://arize.com/docs/PROMPT.md.
Compress documents for LLM token efficiency while preserving semantic content. Use when asked to compress, compact, shrink, or optimize a document, CLAUDE.md, system prompt, skill file, or any text for fewer tokens. Also use when the user mentions token count, token budget, context window limits, or wants to make prompts shorter for cost savings.
Develop, debug, and optimize SGLang LLM serving engine. Use when the user mentions SGLang, sglang, srt, sgl-kernel, LLM serving, model inference, KV cache, attention backend, FlashInfer, MLA, MoE routing, speculative decoding, disaggregated serving, TP/PP/EP, radix cache, continuous batching, chunked prefill, CUDA graph, model loading, quantization FP8/GPTQ/AWQ, JIT kernel, triton kernel SGLang, or asks about serving LLMs with SGLang.
Benchmark any agent skill to measure whether it actually improves performance. Use when the user wants to evaluate, test, or compare a skill against baseline, or when they mention "benchmark", "eval", "skill performance", or "does this skill help". Runs isolated eval sessions with and without the skill, grades outputs via layered grading (deterministic checks + LLM-as-judge), analyzes behavioral signals, and generates a comparison report with a USE / DON'T USE verdict.