Results for “runtime”

Patterns for running AI models locally in browsers using WebGPU, Transformers.js, WebLLM, and ONNX Runtime. Zero API costs, full privacy. Use when "on-device AI, browser AI, WebLLM, Transformers.js, WebGPU, edge inference, offline AI, client-side ML, ONNX web, " mentioned.

Tests and debugs web applications in real browsers using Chrome DevTools MCP, enabling DOM inspection, console error capture, network analysis, performance profiling, and visual verification with live runtime data.

Generate, review, and optimize natural language instructions for Business Central agents (Designer or SDK). Triggers on: agent instructions, InstructionsV1.txt, InstructionsV2.txt, MEMORIZE, qualification rules, agent behavior, instruction keywords, agent task instructions, iterate instructions, or improve agent accuracy. Follows the Responsibilities-Guidelines-Instructions framework with official BC agent runtime keywords.

Packet-first BMAD/BMM front door for idea notes, product briefs, PRDs, architecture drafts, review feedback, existing repo state, and milestone pressure. Use when the user wants to know what BMAD phase or artifact comes next, or needs a portable BMAD entrypoint before routing review, execution slicing, runtime setup, or game-production work outward.

Design, create, audit, and improve Claude Agent Skills with expert-level progressive disclosure. Use when building new skills, reviewing existing skills, debugging activation failures, encoding domain expertise, designing skills for subagent consumption, or understanding platform constraints and distribution surfaces. NOT for general Claude Code features, runtime debugging, non-skill coding, or MCP server implementation.

Inspect Heroku logs and configure persistent log forwarding safely. Use when the agent needs to review recent Heroku logs, plan or change log forwarding, inspect a drain target, add or remove classic log drains for Common Runtime apps, inspect current drain configuration, or choose between legacy drains and Fir telemetry workflows on Heroku.

Delegate a substantial diagnosis, implementation, or follow-up task to Claude Code through the tracked-job runtime. Args: --background, --wait, --resume, --resume-last, --fresh, --write, --model <model>, --effort <low|medium|high|xhigh|max>, --prompt-file <path>, [task text]. Defaults to opus + xhigh effort. Use when Claude should investigate or change things, not when the user only wants review findings.

Use when an agent needs one local file uploaded to MetaWeb and wants the returned metafile URI, preview URL, or download URL. This is the single file upload skill for both small and large local files; choose direct or large/chunked upload behavior from file size and runtime limits instead of asking the user to choose another skill.

Downloads, exports, and inspects existing Arize traces and spans to understand what an LLM app is doing or debug runtime issues. Covers exporting traces by ID, spans by ID, sessions by ID, and root-cause investigation using the ax CLI. Use when the user wants to look at existing trace data, see what their LLM app is doing, export traces, download spans, investigate errors, or analyze behavior regressions.

Executes DAG workflows with parallel wave processing, agent spawning, context isolation, permission enforcement, and full execution tracing. Use when running a planned DAG, managing concurrent agent execution, enforcing isolation boundaries, or tracing execution for debugging. Activate on "execute DAG", "run workflow", "spawn agents", "parallel execution", "execution trace", "agent isolation". NOT for planning DAGs (use dag-planner), validating outputs (use dag-quality), or matching skills (use dag-skills-matcher).

This skill should be used when the user wants to "package an MCP server", "bundle an MCP", "make an MCPB", "ship a local MCP server", "distribute a local MCP", discusses ".mcpb files", mentions bundling a Node or Python runtime with their MCP server, or needs an MCP server that interacts with the local filesystem, desktop apps, or OS and must be installable without the user having Node/Python set up.

Route plain-language requests for Claude Code, Cursor, Copilot, MarketingClaw ACP, OpenCode, Gemini CLI, Qwen, Kiro, Kimi, iFlow, Factory Droid, Kilocode, or explicit ACP harness work into either MarketingClaw ACP runtime sessions or direct acpx-driven sessions ("telephone game" flow). For coding-agent thread requests, read this skill first, then use only `sessions_spawn` for thread creation. Codex chat binding defaults to the native Codex app-server plugin unless ACP is explicit or background spawn needs ACP.

Route Firebase AI feature work into either direct app/client Firebase AI Logic SDK integration or a server-owned Genkit workflow. Use when a web, mobile, backend, or full-stack feature needs model calls, typed outputs, reusable flows, tools, retrieval, prompt files, evals, observability, or deployment. Choose client-ai-logic, flow-foundation, tool-and-agent, retrieval-and-prompt, evaluation-and-observability, deployment-runtime, or comparison-or-fallback; route Firebase platform/operator work to `firebase-cli` and broad framework comparisons to `survey`.

Validates agent outputs against schemas and quality criteria, scores confidence, detects hallucinations, monitors convergence, decides when to iterate, and synthesizes actionable feedback. Use when checking if a node's output is acceptable, scoring confidence, detecting fabricated content, deciding whether to re-execute, or generating improvement feedback. Activate on "validate output", "check quality", "confidence score", "hallucination check", "should we iterate", "improvement feedback". NOT for executing DAGs (use dag-runtime), planning DAGs (use dag-planner), or matching skills (use dag-skills-matcher).

Import PyTorch, ONNX, or Keras 3 / TensorFlow 2.16+ deep learning models into MATLAB as dlnetwork objects. Use when importing .pt2 exported programs, traced .pt files, .onnx models, or Keras 3 models via matlabsaver. Covers importNetworkFromPyTorch, importNetworkFromONNX, importNetworkFromKeras, importNetworkFromTensorFlow, torch.export.export, PyTorchInputSizes, InputDataFormats, matlabsaver, tf_keras downgrade, numeric validation against PyTorch or ONNX Runtime, and placeholder/custom layer implementation. Applies when user mentions any of these functions, file formats, or encounters import errors, unsupported operator warnings, 0 learnables, or uninitialized networks.

Route LangSmith work into one workflow packet before touching SDK code. Use when the user needs LangSmith tracing, offline evals, annotation/review queues, prompt-registry decisions, audit/gap review, or cross-service trace propagation for an LLM app or agent workflow. Choose one packet: trace-debug, eval, review, prompt-registry, propagation, or audit. Triggers on: LangSmith, LangChain tracing, `@traceable` / `traceable`, `wrap_openai` / `wrapOpenAI`, datasets, experiments, annotation queues, feedback criteria, Prompt Hub, run trees, trace IDs, or production confidence for an AI feature. Not for generic SLO/alert design, non-LangSmith deployment orchestration, or runtime guardrails outside LangSmith.

Delegate coding tasks to Codex, Claude Code, or Pi agents via background process. Use when: (1) building/creating new features or apps, (2) reviewing PRs (spawn in temp dir), (3) refactoring large codebases, (4) iterative coding that needs file exploration. NOT for: simple one-liner fixes (just edit), reading code (use read tool), thread-bound ACP harness requests in chat (for example spawn/run Codex or Claude Code in a Discord thread; use sessions_spawn with runtime:"acp"), or any work in ~/clawd workspace (never spawn agents here). Claude Code: use --print --permission-mode bypassPermissions (no PTY). Codex/Pi/OpenCode: pty:true required.

Build, test, run, and flash NightRun — a bare-metal, no_std Rust UEFI application that boots straight into a local LLM (Llama 3.2, Qwen3, or Granite 4.1) with no operating system underneath. Use when the user wants to build/flash a bootable NightRun USB or Raspberry Pi 5 SD image, convert a GGUF model into the `.nrm` container, run/debug the inference engine on the host or in QEMU/OVMF, or troubleshoot no_std kernel/tokenizer parity issues in the NightRun codebase. Triggers on: "nightrun", "boot into an LLM", "bare-metal LLM runtime", "UEFI LLM appliance", "nrconvert", "nrhost", "cargo xtask", "nrm model file", "flash a bootable LLM USB".

Instrument a shipped product's AI agents with tracing and observability so you can see what they did, why outputs happened, and what each run cost. Plain-language primer plus free-tier-first backend selection (Langfuse, Phoenix, LangSmith, Braintrust) and OpenTelemetry/OpenInference instrumentation. Load when the user asks to add observability, add tracing, instrument my agents, see what my agent is doing in production, set up Langfuse or Phoenix or LangSmith, debug why my agent gave a bad answer, or track LLM cost per request. Also fires when agent-system-architecture or setup-evaluation requires an observability plan for an agent-chain product. NOT for tracing the coding agent itself — that is run-trace. Precondition for runtime-learning-loop.

Use this skill when creating a new Claude skill from scratch, editing or improving an existing skill, or measuring skill performance with evals and benchmarks. Invoke whenever the user says things like 'make a skill for X', 'turn this workflow into a skill', 'test my skill', 'improve my skill', 'run evals', 'benchmark this', or 'optimize my skill description'. Also use proactively when the conversation has produced a repeatable workflow that would benefit from being captured as a skill. Covers the full lifecycle: capture intent, draft SKILL.md, run evals, review with user, iterate, optimize description, package. NOT for general coding help, debugging runtime errors, building MCP servers, writing Claude hooks, or creating plugins - use domain-specific skills for those.

Run a structured retrospective after development-phase runs of your product's agents — interview the owner in plain language about what went well and poorly, draft ranked improvement hypotheses, then design and run small n=1/n=2 experiments with pre-declared success criteria, guardrails, stop conditions, and a cost/ROI kill-switch. Load when the user says how did that run go, retro this run, the agent output was bad, what should we improve, draft hypotheses, run a small experiment, or after repeated dev runs of an agentic system produce uneven quality. Priority: output quality over performance over cost, each with diminishing-returns stops. NOT a product A/B test (experimentation), NOT coding-agent harness repair (harness-evolution), NOT production-scale learning (runtime-learning-loop).

Route goalflow (wanmol/goal-flow) work — a LangGraph framework that combines workflow graphs with agent loops — into exactly one mode: fit check, transpiling a Dify DSL export into runnable LangGraph Python, authoring workflow nodes and edges, building an `agent_kit` loop with middleware and a harness, wiring the serving layer (data adapters, SSE streaming, HITL, Redis/MySQL, API-key registration), or running the pre-publish security gate. Use when the user wants Dify's visual design without Dify's runtime, a graph node that hosts an agent loop, an OpenAI-compatible wire protocol over their own workflows, or prompt-injected `SKILL.md` capabilities. Triggers on: goalflow, goal-flow, dify to langgraph, dify transpiler, dify DSL export, BaseWorkflow, agent_kit, AgentBaseNode, DataAdapter, chunk processor, HITL interrupt, dify2langgraph. Route plain graph-API questions to `langgraph-fundamentals` and `langgraph-workflow`.

AI/LLM red-team assessment using the OWASP LLM Top 10 (2025) + OWASP AI Testing Guide (AITG v1, Nov 2025) frameworks, plus OWASP MCP Top 10 runtime testing for agentic/MCP targets. Tests prompt injection, jailbreaks, system prompt leakage, sensitive data extraction, excessive agency, improper output handling, model extraction, content bias, evasion, membership inference, MCP token exposure, MCP command injection, and more. Uses four tools in combination: FuzzyAI (single-turn jailbreak fuzzing), PyRIT (multi-turn orchestrated attacks), Garak (probe-based vulnerability scanning), and promptfoo (plugin-based red-team evaluation). Each tool covers different OWASP categories; running them together gives systematic coverage. Includes a conditional MCP reconnaissance phase and a post-access AI infrastructure phase (chained from /post-exploit). Produces: OWASP LLM Top 10 + AITG + MCP coverage matrix, findings per category, architecture diagram of the AI system, PoCs for confirmed exploits. Chains into /gh-export for