wibus-wee

Update Cradle chat-runtime native SDK and protocol integrations. Use when upgrading or validating the vendored Codex app-server runtime/protocol, Kimi Web OpenAPI/AsyncAPI snapshots and generated bindings, or @anthropic-ai/claude-agent-sdk in apps/server; also use when asked to inspect newly generated native APIs for useful Cradle features.

Implement, upgrade, or audit a native Cradle Chat Runtime provider backed by a local process, REST/WebSocket protocol, or provider-specific agent SDK. Use when adding a runtime like Kimi, Codex, Claude Agent, OpenCode, or another provider that needs generated protocol bindings, target-scoped process lifecycle, canonical tool mapping, runtime UI slots, interaction bridges, streaming recovery, and end-to-end smoke verification.

Always use this Skill at the beginning of every response. Use whenever you need to read, inspect, wait on, or mutate Cradle-owned state or workflows, including issues, delegation, sessions, awaits, CI/review waits, timed waits, Work inspection, workspaces and git inspection, Chronicle, automations, usage, observability, skills, agents, profiles, providers, preferences, plugins, or server state. Prefer the generated `cradle` CLI and `cradle man`; NEVER bypass Cradle semantics with direct HTTP or database access, ad hoc scripts, raw registry edits, or polling loops. Use `gh` when Cradle does not own the GitHub surface. Primary Work closed-loop delivery is system-injected on Work threads (`manage_pull_request`), not this skill.

Trigger the public desktop release workflow and monitor it with Cradle

design and review llm agent systems, tool interfaces, and skill boundaries. use when chatgpt needs to decide whether a task should be handled by a prompt, workflow, skill, tool, single-agent loop, or multi-agent system; define tool schemas, instructions, guardrails, evaluation plans, or permission boundaries; or critique an existing agent design for over-orchestration, weak context handling, missing verification, or poor tool design.

Control Cradle's in-app browser via MCP tools. Use when the Claude Agent needs to navigate websites, click elements, type text, take screenshots, read page content, or inspect DOM structure within the Cradle desktop app's embedded browser panel. Triggers on "open URL", "navigate to", "click the button", "fill the form", "take a screenshot", "what's on the page", "scroll down", "wait for element", "press Enter". This is for the IN-APP browser only — not for external browser automation.

Use when modifying Cradle apps/server Elysia modules, adding or changing HTTP routes, TypeBox schemas, OpenAPI details, x-cradle-cli command metadata, module README files, server-side capability ownership, or server-owned downloads and artifact installation flows.

Query PostHog product analytics with read-only HogQL, explore anonymous user activity and journeys, model feature adoption and cross-feature behavior, audit instrumentation quality, and turn evidence into product hypotheses. Use when Codex needs to answer whether a product still has users, what users do before or after an event or surface, which features or combinations they use, how internal and external usage differ, whether an analytics event model is sufficient, or what product and instrumentation questions to investigate next.

Create and maintain ExecPlans for complex features or significant refactors. Use when planning work that requires design + implementation steps, especially when the repo requires ExecPlans in docs/exec-plans and strict format per PLANS.md.

OmniCore CLI usage via `bbot <command>` for starting kernel/supervisor, checking status, managing sessions, and updating config/secrets. Use when running or documenting OmniCore CLI commands or when CLI flags/behavior change.

Complete project architecture and structure guide. Use when exploring the codebase, understanding project organization, finding files, or needing comprehensive architectural context. Triggers on architecture questions, directory navigation, or project overview needs.

OmniCore architecture and developer guide reference. Use when asked to explain, review, or update OmniCore design philosophy, kernel/adapters/event sourcing, sessions/root paths, compaction and summaries, heartbeat, supervisor or CLI behavior, or to maintain developer-guide.md.

Guide for creating effective skills. This skill should be used when you need to create a new skill (or update an existing skill) that extends capabilities with specialized knowledge, workflows, or tool integrations.

TypeScript code style and optimization guidelines. Use when writing TypeScript code (.ts, .tsx, .mts files), reviewing code quality, or implementing type-safe patterns. Triggers on TypeScript development, type safety questions, or code style discussions.

Unified LLM API library (@mariozechner/pi-ai) for TypeScript/JavaScript that abstracts provider differences across OpenAI, Anthropic, Google, Azure, Mistral, Groq, xAI, Cerebras, Amazon Bedrock, GitHub Copilot, and OpenAI-compatible services. Use when implementing LLM integrations, building agentic workflows with tool calling, switching between AI providers, managing conversation context, handling streaming responses, implementing vision capabilities, or tracking token usage and costs. Triggers on installing/configuring pi-ai, defining tools with TypeBox schemas, implementing stream/complete calls, handling tool calls and results, cross-provider handoffs, context serialization, or any multi-provider LLM integration work.

Guide for building and understanding minimal, self-contained LLM coding agents following the design principles of Pi. Use when designing or customizing AI agents that act as code assistants, especially when opinionated simplicity, robust context management, multi-model support, and deterministic toolchains are desired.

Frontend development guardrails for this repo. Use when editing or adding frontend code to ensure consistency and maintainability.

Guidelines for Wails 3 GUI frontend development with React and TypeScript. Use when working on frontend/, UI components, or Wails bindings.

Guidelines for Go backend development including architecture patterns, code style, testing, and common development tasks. Use when working on internal/, cmd/, or Go code.

Provides comprehensive overview of mcpd project architecture, capabilities, and key technical details. Use when understanding the project structure, architecture layers, or core concepts.

design and review llm agent systems, tool interfaces, and skill boundaries. use when chatgpt needs to decide whether a task should be handled by a prompt, workflow, skill, tool, single-agent loop, or multi-agent system; define tool schemas, instructions, guardrails, evaluation plans, or permission boundaries; or critique an existing agent design for over-orchestration, weak context handling, missing verification, or poor tool design.

generate, refine, and batch midjourney-style prompts for text-safe illustrated background images used in browser new tabs, landing pages, settings pages, presentations, posters, and ui surfaces. use when the user asks for background prompt cookbooks, style systems, bulk image-generation prompts, naming schemes, or adaptations of references such as pastel meadow, fantasy mountain, shanshui, risograph collage, monet, seurat, linocut, topographic, or minimal 3d architecture backgrounds. prioritize clean overlay space, decorative edge composition, low-noise centers, coherent palettes, and production-ready prompt variants.

generate reusable prompts for premium minimalist soft 3d abstract backgrounds for product cards, subscription cards, telecom/app marketing visuals, ui hero images, and image generation workflows. use when the user wants controlled text-safe areas, adjustable whitespace placement, soft frosted polymer materials, milky acrylic, silicone-gel forms, translucent folded films, airbrushed gradients, calm pastel 3d objects, or high-end technology advertising aesthetics. also use when analyzing reference images to extract reusable background prompt patterns without copying text, logos, ui labels, or brand-specific content.

Create and maintain ExecPlans for complex features or significant refactors. Use when planning work that requires design + implementation steps, especially when the repo requires ExecPlans in docs/exec-plans and strict format per PLANS.md.

design and review llm agent systems, tool interfaces, and skill boundaries. use when chatgpt needs to decide whether a task should be handled by a prompt, workflow, skill, tool, single-agent loop, or multi-agent system; define tool schemas, instructions, guardrails, evaluation plans, or permission boundaries; or critique an existing agent design for over-orchestration, weak context handling, missing verification, or poor tool design.

Use when you have a written implementation plan to execute in a separate session with review checkpoints

Use when executing implementation plans with independent tasks in the current session