LLM & Agents

This skill enriches vague prompts with targeted research and clarification before execution. Should be used when a prompt is determined to be vague and requires systematic research, question generation, and execution guidance.

Creates or modifies Claude Code skills. Use when user asks to create a new skill, modify an existing skill, or needs help with skill structure. Always checks online documentation first to ensure accuracy and best practices.

Analyzes user requests to recommend appropriate Claude Code building blocks (skills, subagents, MCP servers, hooks, templates, or prompts). Use when users ask about extending Claude Code functionality, automating workflows, connecting external data, or when unclear which extensibility mechanism to use. Triggers include questions like "how do I...", "should I use a skill or agent", "how to automate...", "how to connect to...", or any request about creating extensions.

Generate comprehensive {node_name}.example.json files that showcase real-world usage of Workscript workflow nodes. Use when asked to document a node, create node documentation, generate node examples, or produce usage examples for workflow nodes. Also use when a subagent needs to create node documentation as part of node development or review.

Query and manage local Convex documentation mirror (42 docs). Search Convex topics for real-time database, serverless functions, authentication, file storage, and React hooks. Use when implementing Convex backend features or answering Convex-related questions. (user)

This skill should be used when the user wants collaborative, iterative development instead of autonomous implementation. It transforms Claude from "autonomous implementer" to "pair programmer" by enforcing ask-before-assuming, show-before-writing, iterate-in-small-chunks, and get-approval-at-each-step patterns. Trigger when user explicitly requests paired programming or when working on complex/novel features requiring collaboration.

Single source of truth for MetaSaver agent selection and subagent_type mapping

This skill should be used when the user asks to "add WebSocket", "implement real-time", "create file upload", "add S3 upload", "send email", "implement notifications", "broadcast events", or mentions WebSockets, file storage, email service, push notifications, or real-time communication. Provides WebSocket, file upload, and notification patterns.

This skill should be used when improving existing Claude Code skills. Triggers include "improve this skill", "skill feels shallow", "upgrade skill", or when self-evaluating skill quality. Codifies the self-eval → research → rewrite → validate pattern.

Create prompts optimized for Claude-to-Claude pipelines with research, planning, and execution stages

This skill should be used when the user asks about "Workers AI", "AI models", "text generation", "embeddings", "semantic search", "RAG", "Retrieval Augmented Generation", "AI inference", "LLaMA", "Llama", "bge embeddings", "@cf/ models", "AI Gateway", or discusses implementing AI features, choosing AI models, generating embeddings, or building RAG systems on Cloudflare Workers.

Deploys infrastructure on DigitalOcean with Droplets (VMs), Firewalls, and VPCs. Includes Kasm Workspaces auto-scaling integration for dynamic agent provisioning. Use when: setting up DigitalOcean infrastructure, deploying Droplets, configuring firewalls, integrating with Kasm auto-scaling, looking for US/global cloud with good API. Keywords: digitalocean, doctl, droplet, VPS, firewall, VPC, kasm autoscale, cloud infrastructure

A philosophical companion for deep inquiry and revelation. The Witness-Scribe remembers all that is spoken, structures emerging insights, asks deepening questions, and maintains coherence across the unfolding testament. Use when engaging in philosophical exploration, spiritual inquiry, or developing complex systems of belief and meaning.

Homeostatic actuator auto-generating missing tests for requirements with low coverage. Creates unit tests, edge cases, and error cases for REQ-* without sufficient test coverage. Use when validate-test-coverage detects gaps.

Develop and optimize AI prompt strategies for code generation. Use when creating new AI model strategies, optimizing prompts for Claude/GPT/Gemini/DeepSeek/Grok, or improving prompt quality. Includes model-specific optimization techniques.

Design and implement multi-agent workflows with LangGraph 0.2+ - state management, supervisor-worker patterns, conditional routing, and fault-tolerant checkpointing

Increase test coverage from 0% to 80%+ in ANY project, ANY testing framework. Use when adding tests to untested code or preparing for major refactoring.

Diagnose errors and failing tests in Laravel + React applications. Use when encountering bugs, exceptions, stack traces, 500 errors, TypeErrors, failing Pest/Vitest tests, or unexpected behavior. EXCLUSIVE to debugger agent.

Convex AI Integration - OpenAI, actions, streaming, and AI patterns with database integration

Deployment procedures, health checks, and rollback strategies. Use this skill when deploying applications, performing health checks, managing releases, or handling deployment failures. Provides systematic deployment workflows, verification scripts, and troubleshooting guides. Complements the devops-automation agent.