Skills

Connect to InsForge along three paths: CLI, SDK, and OpenRouter

For people who want an agent to build backend apps with InsForge. Install first with the official skills add. Backend uses CLI skills, clients use the SDK, new AI features need ai setup first, and diagnostics only locate issues without giving fixes.

Official source · insforge-skills

Install

npx -y skills add insforge/insforge-skills

Who it's for

Use this skill set when you want an AI agent to build apps with InsForge, a Backend-as-a-Service. The source writes that they help developers build apps through an agent. Product page:

https://insforge.dev/

Skill repository:

https://github.com/insforge/insforge-skills

The source gives one key distinction: backend infrastructure uses CLI skills; most client integrations use @insforge/sdk; new AI features use the OpenRouter API, with keys configured by npx -y @insforge/cli ai setup. If you are not ready to run the official install on this machine, or you do not accept writing skill files, do not continue yet.

There is also a diagnostics skill: it guides running diagnostic commands to locate problems and does not provide fix suggestions. Do not expect it to change code directly. The source does not expand backend capabilities into a feature inventory; do not invent concrete business operations. If you only want to click InsForge in a browser and are not ready to let an agent wire via CLI skills or the SDK, you need not start here.

First try in 3 minutes

First open the official repository and check:

https://github.com/insforge/insforge-skills

Recommended install is only this one. Installing skills changes the local skill directory; before you run it, confirm you want InsForge skills left here:

npx -y skills add insforge/insforge-skills

Success signal: this command ends normally. The source does not write a skill landing path; do not invent a directory. The source also has plugin-marketplace installs aimed at particular clients; the first time, do not put them beside this as a second main path, and do not apply marketplace commands to unexplained environments.

The first time in the agent, only check; do not run ai setup yet, and do not invent diagnostic subcommands. Ask the agent to recite the key distinction: backend infrastructure takes CLI skills, client integration mostly uses @insforge/sdk, new AI features take OpenRouter, and diagnostics only locate problems without giving fix suggestions. The observation is that it can state these three paths clearly, has not configured keys, and has not run unwritten repairs.

ai setup configures an API key and is a side effect. Run it only when you confirm you want an OpenRouter key ready for new AI features:

npx -y @insforge/cli ai setup

The source does not write a key field name; do not write a placeholder key. The first time, do not treat the diagnostics skill as auto-repair. If the agent offers to run a diagnostic command that never appeared, first have it state that the source only writes guidance for locating issues and gives no command inventory, then stop.

Use cases

Case 1: First install the skills via the official entry. In an agent that can receive skills add, after you accept writing skill files, only run npx -y skills add insforge/insforge-skills. Fits people who only want InsForge skills attached to an agent and are not ready to configure keys. Success signal: the install command ends normally, and the agent can say this is for InsForge Backend-as-a-Service. If install fails, stop; do not switch to the plugin-marketplace path as a backup main path.

Case 2: First check the wiring, then decide whether to configure an AI key. After install, have the agent recite: backend infrastructure uses CLI skills, most client integrations use @insforge/sdk, new AI features use the OpenRouter API, keys are configured by npx -y @insforge/cli ai setup; the diagnostics skill only locates problems and does not provide fix suggestions. Fits people who worry the agent will go online or change the backend at random. Success signal: the three-path distinction and the diagnostics boundary are both stated clearly. Before you really run ai setup, confirm you want the key left in place. The source gives no diagnostic command inventory; do not invent one. Do not treat "help developers build apps" as already giving concrete create or migrate steps.

Boundaries

The main path is only npx -y skills add insforge/insforge-skills. The source also wrote plugin-marketplace install; that is not the first-try main path. Do not put them side by side. Product page:

https://insforge.dev/

The three-path distinction must hold: backend infrastructure takes CLI skills, client integration mostly uses @insforge/sdk, new AI features take OpenRouter and finish ai setup first. The diagnostics skill only guides locating problems and does not provide fix suggestions. The source does not expand a concrete backend-operation inventory, and it does not write the skill directory, key fields, or uninstall steps. Do not fill in unwritten business actions, repair commands, or a client roster. First-time success is only install finishing and the agent reciting the distinction; before configuring a key or running diagnostics you must confirm the side effects. If the agent starts giving repair steps, stop first and return to "locate problems only". Do not pretend OpenRouter is already wired for new AI features before ai setup finishes. Do not bypass @insforge/sdk for client integration and invent another SDK name. Do not switch backend infrastructure to an installer that never appeared. Evidence does not write uninstall or update paths; do not invent commands. First-time success only requires install finishing, plus the agent stating the three paths of CLI skills, SDK, and OpenRouter clearly.