LLM 與 Agent

MANDATORY skill when KeyError or schema errors occur. Fetch actual prompt schemas instead of guessing. Use for debugging traces and understanding AI model behavior.

Structured code debugging through hypothesis formation and falsification planning. Use when diagnosing bugs, unexpected behaviour, or system failures where the root cause is unclear. Produces a hypothesis document for execution by another agent rather than performing the investigation directly. Triggers on requests to debug issues, diagnose problems, investigate failures, or create debugging plans.

Génère une nouvelle complète style Twilight Zone (8000-15000 mots) EN FRANÇAIS inspirée par un sujet scientifique/technologique. Utilise ce Skill pour créer une fiction autonome avec twist inattendu, exploration philosophique de la condition humaine, et conclusion réflexive dans l'esprit authentique de Rod Serling. Référence obligatoire: twilight_zone_episodes.json.

Build Gemini Deep Research agents using the Interactions API. Use when creating autonomous research agents, implementing multi-step information gathering, building research report generators, or integrating Gemini's Deep Research capabilities into applications. Handles polling, streaming, file integration, and report synthesis.

OpenAI API via curl. Use this skill for GPT chat completions, DALL-E image generation, Whisper audio transcription, embeddings, and text-to-speech.

Enterprise Mermaid diagramming skill for Claude Code using MCP Playwright. Use when creating architecture diagrams, flowcharts, sequence diagrams, or visual documentation.

This skill should be used when setting up, managing, or optimizing Grail miners on Bittensor Subnet 81. Use it for GRAIL protocol tasks including miner setup, R2 storage configuration, model checkpoint management, GRPO rollout generation, performance optimization, competitive monitoring, and troubleshooting common issues like CUDA errors, upload failures, or low scores. Essential for miners working with verifiable post-training, SAT/GSM8K environments, or understanding the GRAIL incentive mechanism to improve competitiveness.

New agent startup checklist for Agent Mail and Beads. Use when starting a new agent session, when the user says "prime" or "startup", or when beginning work on a multi-agent project.

LLM evaluation and prompt optimization with Evidently.ai. Covers textdescriptors, dataset metrics, LLM-as-a-Judge patterns, and automatedprompt optimization for classification and generation tasks.

Use when facing 3+ independent failures that can be investigated without shared state or dependencies. Dispatches multiple Claude agents to investigate and fix independent problems concurrently.

Build mithril-cache for torch.compile caching. Use when implementing content-addressable storage, cache keys, eviction, or framework hooks.

Automatically adds timestamps and execution duration to all Claude responses

Expert on integrating Shelby Protocol decentralized storage into applications. Helps with use case evaluation, architecture design, migration strategies, cost optimization, performance tuning for video streaming, AI training, data analytics, content delivery, and read-heavy workloads. Triggers on keywords integrate Shelby, decentralized storage integration, video streaming storage, AI training data, data analytics storage, migration to Shelby, storage architecture, content delivery, Shelby use case.

Coordinate with other AI agents running in tmux panes. Use this skill when you need to delegate tasks, request reviews, or collaborate with agents like Codex, Gemini, or other Claude instances.

Create or resume an execution plan - a design document that a coding agent can follow to deliver a working feature or system change

MANDATORY protocol for executing implementation plans with independent tasks - dispatches fresh subagent for each task with code review between tasks, enabling fast iteration with quality gates. MANDATORY parallel execution for independent tasks.

指定したプラグインをインストールする。「プラグインをインストール」「〇〇を追加して」「プラグインを入れて」「〇〇をインストールして」「プラグインを導入」「〇〇を使えるようにして」「新しいプラグインを追加」などで起動。claude plugin install コマンドを使用してインストール。

Run and manage functional tests (unit, integration, E2E, mutation). Use when running tests, debugging test failures, ensuring test coverage, or fixing mutation testing issues. Covers PHPUnit, Behat, and Infection. For K6 load/performance tests, use the load-testing skill instead.

This skill should be used when the user asks to "create mermaid diagram", "add visuals to slides", "suggest stock photos", "generate AI image prompts", "create diagram for presentation", "design visual theme", or needs guidance on diagram types, color palettes, image selection, or AI image generation prompting for presentations.

Comprehensive guide for building Agent Skills according to agentskills.io specification. Use when developers need to create, structure, or validate skills that follow the Agent Skills standard format and best practices.