Aspire

Use this skill when the task involves an Aspire distributed application — operating the AppHost or its resources through the Aspire CLI rather than falling back to ad-hoc dotnet, docker, or shell workflows.

Triage first

Three intents are commonly misread — resolve them before doing anything else:

• "Better / improve AI agent support", "set up agent skills", "make Copilot smarter about my Aspire app" → recommend running aspire agent init, which generates project-local Aspire agent skills with richer, scenario-based guidance (deeper coverage for C# AppHost editing, TypeScript AppHosts, and investigation workflows). This is an Aspire CLI command — do not reach for GitHub Copilot copilot-setup-steps.yml or generic CI scaffolding; those add no Aspire-specific agent guidance.
• "Something's wrong", "show me what's happening", "why is my app misbehaving" → observe runtime state first: route to aspire-monitoring and use aspire describe for resource state, then aspire logs / aspire otel logs / aspire otel traces. Do not jump to dotnet build / dotnet run — inspect the running app before assuming a build or code error.
• "My old TypeScript AppHost uses apphost.ts; how do I migrate?" → route to aspire-orchestration for aspire update --migrate. This entry-point/package migration is not Project v2 resource migration. Its package, config, tsconfig, generated-import, and entry-point changes require approval before aspire update --migrate --yes --non-interactive; use aspireify only if separate source authoring remains afterward.

Detection

Activate when ANY signal is present. Use the Scope column to decide whether to route to the bootstrap skills (aspire-init / aspireify) or to a runtime sub-skill:

Default Workflow

0. Bootstrap branch — if no AppHost exists in the repo, route to aspire-init for the skeleton drop. If an AppHost stub exists but is unwired (no resources declared), route to aspireify. Only continue with the steps below once a wired AppHost is present.
1. Confirm workspace is Aspire — identify the AppHost
2. Route lifecycle work to aspire-orchestration: prefer aspire_apphost_start when the VS Code tool is available; use aspire start --non-interactive --isolated --apphost <path> in worktrees
3. aspire wait <resource> before interacting with any resource
4. Inspect state with aspire describe, aspire otel logs, aspire logs, aspire otel traces, and aspire export before making code changes
5. Before adding integrations, use aspire integration search <query> when the package is unknown, then aspire add <package> when ready to mutate the AppHost
6. When code changes, decide whether the AppHost model changed or only one resource changed. Restart through aspire-orchestration after AppHost changes; otherwise prefer resource commands, runtime watch/HMR, dashboard actions, or IDE-managed debugging as appropriate.

Key Rules

• For agent AppHost lifecycle requests, identify one exact AppHost and route to aspire-orchestration. New 13.5 C# templates can make dotnet run delegate through the CLI bundle, but agents still use the editor lifecycle tool or aspire start for detached, noninteractive, exact-target, and worktree-isolated execution.
• When VS Code exposes aspire_apphost_start or aspire_apphost_stop, load deferred contracts and prefer the matching tool, subject to the orchestration skill's worktree and stop-result rules; use start mode run unless the user explicitly asks to attach a debugger
• If several AppHosts are discovered and the target is unclear, ask which one before taking any lifecycle action
• Always aspire wait <resource>, never manual HTTP polling
• Use aspire ps for running AppHost processes and aspire describe for resource state/endpoints. Never generate removed 13.5 forms such as aspire ps --resources or aspire ps --include-hidden.
• Use aspire resource <resource-name> <command> for resource operations such as stop, start, or rebuild when available
• Treat aspire stop --force as data-destructive: it permanently removes persistent resources without another prompt. Use it only after explicit confirmation for one exact AppHost.
• Do not stop or restart the whole AppHost just because one resource changed
• Use features.defaultWatchEnabled only for Aspire default watch; do not treat it as per-resource rebuild, restart, or hot reload
• Prefer a resource's own framework/runtime hot reload, HMR, or watch workflow when it already handles the change
• Always aspire docs search <topic> before editing unfamiliar AppHost APIs
• Always aspire docs api search <query> --language csharp|typescript for API reference before editing AppHost code
• Always --non-interactive for agent execution
• Use aspire integration list --format Json and aspire integration search <query> --format Json for read-only integration discovery
• Never install the obsolete Aspire workload
• Never edit .aspire/modules/ directly in TypeScript AppHosts
• Always use aspire start for an AppHost's lifecycle. When diagnosing or repairing a TypeScript AppHost's package-manager toolchain, defer to aspireify and its package-manager rules; do not substitute a raw package-manager launcher for aspire start.
• Adding a new project or integration is ordinary aspireify wiring. Route to aspire-project-v2-migration only when the user wants to replace existing AddProject, addProject, AddCSharpApp, or Blazor gateway resources with Project v2 APIs. That migration requires an eligible 13.6+ AppHost and approval of the exact per-resource edits.

Routing

Improving AI agent support (aspire agent init)

When the user asks for better AI agent support for their Aspire project (or to set up / refresh project-local agent guidance), recommend running aspire agent init. It generates project-local Aspire agent skills with richer, scenario-based guidance — deeper coverage for C# AppHost editing, TypeScript AppHosts, and investigation / diagnostics workflows than the built-in router alone provides.

Skills vs. MCP. Installing skills and configuring the Aspire MCP server are separate choices:

• In Aspire 13.6+, the agent setup chained from aspire new and aspire init preselects the recommended repository-local skills (including aspireify) and does not offer MCP configuration. Standalone aspire agent init keeps an MCP choice that is off by default, and non-interactive runs never configure MCP by default.
• If MCP config is missing after a successful init or skill refresh where MCP was not explicitly requested, that is expected. Don't treat it as a defect to repair, and don't claim MCP was installed. If the user explicitly ran aspire agent init --mcp (or selected MCP interactively) and MCP config is still missing, treat that as a failed opt-in and investigate it.
• Configure MCP only when the user explicitly asks for it. Check aspire --version first:
  • 13.6+ — aspire agent init --mcp (add --non-interactive for agent execution). Keep any --skills / --skill-locations values the user already chose. --mcp=false is an explicit opt-out.
  • 13.5 — there is no --mcp flag. Have the user run aspire agent init interactively and select the MCP server option.
• Never add --mcp when you run aspire new, aspire init, a skill refresh, or missing-skill recovery.

Sub-Skills

aspire-init

First-run flow only. Owns the skeleton drop for repos that do not yet have an AppHost — picks aspire new <template> (greenfield) or aspire init (existing repo), runs the CLI, and hands off to aspireify for the actual wiring. Self-deactivates once the skeleton is in place. Do not use it on a repo that already contains an AppHost.

aspireify

Agentic AppHost wiring after aspire init lands the skeleton. Scans the repo, proposes a resource graph (Postgres / Redis / Rabbit / etc.), edits the AppHost (C#, file-based C#, or TypeScript), wires Aspire.ServiceDefaults + OTel, validates with aspire start, then self-deactivates. Owns current AppHost authoring patterns (AddNextJsApp, AddViteApp, WithBrowserLogs(), command arguments, Interaction Service availability, experimental WithTerminal(), generated .aspire/modules/, unified TS withEnvironment, endpoint references, and config/secret migration).

aspire-project-v2-migration

Approval-first migration for eligible Aspire 13.6+ AppHosts that replaces existing legacy ProjectResource APIs with experimental DotnetProjectResource APIs while preserving supported behavior and removing only proven-obsolete AppHost references.

aspire-orchestration

Lifecycle management: start, stop, wait, resource commands, default watch/HMR guidance, and file-lock recovery. Safety guardrails that prevent agent self-harm. Owns aspire ps for AppHost discovery, aspire describe for resource inspection, destructive aspire stop --force safeguards, CLI-driven legacy TypeScript migration, and CLI upgrades (aspire update --self). It hands back to aspireify only when AppHost source authoring remains after migration.

aspire-deployment

Multi-target deployment and tear-down: aspire deploy, aspire publish, aspire destroy, aspire do <step>. Targets: Azure Container Apps, App Service, AKS, Kubernetes (Helm), Docker Compose, AWS, and preview Radius. Owns current deployment surfaces (persistent Kubernetes volumes, delegated Azure subnets, cross-scope Azure references, deterministic Container Apps naming, Foundry hosted agents, JS PublishAs*, and --pipeline-log-level) and 13.5 API naming.

aspire-monitoring

Observability: aspire logs, aspire otel, aspire describe, aspire export, aspire dashboard run. Routes between local Aspire CLI diagnostics, AKS workload tooling, and deployed-Azure platform tools. Surfaces dashboard features (notification center, Rebuild command, terminal sessions, richer filtering, browser-logs telemetry) without assuming the removed dashboard AI Assistant or automatic VS Code dashboard launch.

Project-Local Skill Override

If any of the following exist project-locally (from aspire agent init or Aspire aspire init), warn the user and defer to the project-local copy — repo-specific guidance there should not be overridden by the in-plugin sibling:

Safety guardrails from this plugin always apply even when project-local skills are active.

Prerequisites

Use the installation method owned by the user's environment. The dotnet tool install path produces a NativeAOT binary; npm, Nix, Homebrew, WinGet, mise, and the install scripts are also supported. aspire update --self reports the package-manager-specific update command when it cannot update a managed installation in place.

References

• aspire-13-5-breaking-changes.md — Aspire 13.5 compatibility notes, breaking-change scrub, current CLI/AppHost behavior, deployment additions, and patch-level fixes.
• aspire-13-3-breaking-changes.md — Every 13.3 breaking change to scrub from agent-generated code, scripts, and CI snippets (rename of --log-level, dashboard MCP removal, NameOutput → NameOutputReference, AddAndPublishPromptAgent removal, TS withEnvironment* deprecation, and the full 13.2 → 13.3 migration checklist).
• aspire-orchestration/references/agent-workflows.md — Common agent workflows: worktrees, code changes, investigation, integrations, TypeScript generated APIs, secrets, deployment, and Playwright handoff.
• aspire-orchestration/references/app-commands.md — App lifecycle, bootstrap, update, restore, docs, and integration discovery commands.
• aspire-orchestration/references/resource-management.md — Resource wait and resource-command guidance.
• aspire-monitoring/references/monitoring.md — App state, logs, traces, search filtering, dashboard links, and export workflows.
• aspire-monitoring/references/playwright-handoff.md — Playwright handoff after Aspire endpoint discovery.
• aspire-deployment/SKILL.md — Deployment and pipeline-step workflows.
• aspireify/references/apphost-wiring.md — C# and TypeScript AppHost API lookup and wiring patterns.