Drop-in inspector panel for any json-render app. Use when the user wants to debug a generative UI, inspect the spec tree, edit state at runtime, see dispatched actions, follow stream patches live, browse a catalog, or pick DOM elements to find their spec keys. Triggers include "add devtools", "debug json-render", "inspect the spec", "why is this element not rendering", "see the state at runtime", or requests to tap streams / capture action logs for `@json-render/devtools`.

Creates JavaScript classes using Bun's bindings generator (.bind.ts). Use when implementing new JS APIs in Rust with JSC integration, prototypes, or constructors.

Design, build, debug, and optimize OpenRouter API integrations. Use for any OpenRouter API work, including models, reasoning, routing, media, tools, structured output, cost, or performance.

Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.

Manage Meta Ads (Facebook + Instagram) — performance, ROAS, CPM, frequency, audience overlap, learning phase, creative fatigue, budgets, ad sets, campaigns, ads. Use for any mention of Meta Ads, Facebook Ads, Instagram Ads, ROAS, CPM, ad spend, or campaign settings on Meta.

Manage the user's shared agent-skill library via skills-manager-cli — install, update, remove, deploy or undeploy skills per agent, manage presets, organize tags, search, and adopt existing skills. Use this whenever the user wants Claude Code, Codex, Cursor, or another agent to gain or lose a skill, wants to organize the central library, or asks what is installed or deployed. Prefer this over direct agent-folder installs because Skills Manager preserves source metadata, preset membership, updates, and cross-agent deployment state.

Locate objects with a running PerceptionLocalize stack over MCP (`state`, `localize`) - probe the machine, build the time window, watch labels on a schedule, narrow an object's position, change params per call, and tell where a failure comes from. Use whenever a blueprint with PerceptionLocalize and McpServer runs, such as go2-localize or xarm-localize-live.

Set up or modernize code-quality tooling for JavaScript and TypeScript projects. Use when creating a project, adding code-health or CI quality checks, making a repository agent-ready, or consolidating dead-code, duplication, architecture, dependency, and changed-code analysis. Do not use for formatting-only, lint-rule-only, or TypeScript type-error tasks. Do not use to run an analysis of a project that is already set up; use the fallow skill for that.

Activate when the user explicitly names the narrator skill or requests production-ready narration in either case: (1) numbered takes fitted to fixed video windows, or (2) one continuous long-form story read in a locked voice that must be measured and retried for timbre or internal pauses. Collect missing text, voice_id, or voice_type after activation; their absence is not a reason to skip this workflow. Own pacing, speech-duration gates, retries, and ready audio delivery; never time-stretch. Do not use for ordinary TTS, "read this aloud", an unconstrained voiceover, voice browsing, music, SFX, singing, dubbing existing speech, or voice cloning.

Share Dropbox files or folders by creating shared links, inspecting existing shared links, or explaining current shared-link state. Use only when the user explicitly asks to create a shared link, check access, or inspect a Dropbox shared link.

Async and concurrency patterns for Mistral Vibe. Use when working with asyncio, the agent loop, the app server's event loop, HTTP clients, streaming surfaces, or core-to-client communication.

Analyze MSBuild binary logs to diagnose build failures. USE FOR: build errors that are unclear from console output, diagnosing cascading failures across multi-project builds, tracing MSBuild target execution order, and generally any MSBuild build issues. Requires an existing .binlog file. DO NOT USE FOR: generating binlogs (use binlog-generation), binlog MCP startup or connection failures (use troubleshoot-binlog-mcp), non-MSBuild build systems.

Correctly call native (C/C++) libraries from .NET using P/Invoke and LibraryImport. Covers function signatures, string marshalling, memory lifetime, SafeHandle, and cross-platform patterns. USE FOR: writing new P/Invoke or LibraryImport declarations, reviewing or debugging existing native interop code, wrapping a C or C++ library for use in .NET, diagnosing crashes, memory leaks, or corruption at the managed/native boundary. DO NOT USE FOR: COM interop, C++/CLI mixed-mode assemblies, or pure managed code with no native dependencies.

The scheduled model sweep. Reads the team's models channel for hints, discovers new LLM models, adds them to ml_model_list.py, runs the paid tests, marks dead provider entries deprecated, opens PRs under the team's rules, announces them in Slack, and checks whether remote_config needs a publish PR. Use when asked to run the model sweep by hand, to add or deprecate models unattended, or to change how the daily sweep behaves.

Kiln's code conventions and per-area gotchas, plus a diff gate that enforces the mechanical ones. Invoke before writing or changing any Python, TypeScript or Svelte code in the Kiln repo (libs/core, libs/server, app/desktop, app/web_ui), and when reviewing such a change. Covers comments, startup and entry points, globals and module state, config reads, module layering, library-vs-app rules and async I/O.

Create, migrate, update, and verify Wear OS widgets using Glance Wear and Remote Compose in an Android project. Use this skill when implementing GlanceWearWidget, GlanceWearWidgetService, WearWidgetDocument, Remote Compose UI primitives (RemoteColumn, RemoteRow, RemoteBox, RemoteText, RemoteImage, RemoteButton), local state with rememberMutableRemoteInt and valueChange, click actions with pendingIntentAction, widget provider XML containers (SMALL and LARGE), rectangular widget picker PNG previews, or migrating legacy ProtoLayout TileService tiles to Wear OS widgets. Don't use for mobile home screen AppWidgetProvider or GlanceAppWidget implementations, watch face complications, or full-screen Wear OS activities.

Download and archive Discord attachments, images, and links from a server channel, DM, or Discord Data Package export. Use when asked to "download Discord attachments", "harvest a Discord channel", "archive a DM", or "organize a Discord Data Package"; or when preserving files from a Discord export or conversation.

Use the local Kaggle CLI skill for command guidance, workflows, and troubleshooting across competitions, datasets, kernels/notebooks, models, model variations and versions, inbox file uploads, forums/discussions, benchmarks, configuration, OAuth/API-token authentication, and accelerator quota. Activate this skill when the user asks about kaggle CLI commands, examples, flags, metadata files, download/upload flows, submissions, benchmark tasks, or Kaggle CLI behavior.

Manually dispatch and iterate on the Release Canary workflow that smoke-tests published OpenShell artifacts (install.sh on macOS/Ubuntu/Fedora, Helm chart on kind) after each Release Dev publish. Use when changing `.github/workflows/release-canary.yml`, validating a release before tagging, debugging a canary failure, or reproducing a canary job locally. Trigger keywords - release canary, release-canary, canary failed, canary dispatch, test release canary, post-release smoke, install.sh canary, helm chart canary, kind canary, dispatch canary.

Reach into past AI conversation history through the `trellis mem` CLI. Use whenever the user asks 'how did we solve X last time', 'have we discussed this before', 'what was the decision on X', 'remind me what we did in this task', '上次怎么解的', '之前讨论过吗', '想起一段对话', or when starting a brainstorm that overlaps prior work, debugging a familiar bug, continuing a task across sessions, or doing a finish-work review. Returns raw past dialogue; decide for the moment whether to update spec, append to task notes, quote inline in the answer, or just internalize.