# Run Cloud AI Apps from the Command Line

Cloud AI services often require separate APIs and workflows. This skill guides Claude, Codex, and Claude Code through inference.sh app discovery and execution.

## Install

```bash
npx skillstore add 101-skills/infsh-cli
```

## Metadata

- Status: approved
- Slug: 101-skills-infsh-cli
- Skillstore revision: r1
- Version status: missing
- Tree hash: e5d60be87d2859c521a2de6e7f9f01c6598452fb1a4a89d097315ed4154de378
- Author: 101-skills
- GitHub username: 101-skills
- License: MIT
- Repository: https://github.com/101-skills/skills/tree/main/tools/infsh-cli/
- Ref: d71c7417a35d5c2624161bd2fe8de8a41a362128
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: blocked
- Manual install advisory: allowed\_with\_warning
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, env\_access, filesystem, external\_commands
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/101-skills-infsh-cli
- Manifest: https://skillstore.pages.dev/api/skills/101-skills-infsh-cli/manifest

## Capabilities

- Search and filter the public inference.sh app store.
- Inspect app details and required input fields.
- Generate sample input files for available apps.
- Run image, video, audio, text, search, 3D, and social apps through belt.
- Submit long tasks asynchronously and retrieve their status or results.
- Upload local media automatically when an app input receives a file path.

## Use Cases

- Create Media Prototypes: Find suitable image or video apps, inspect their inputs, and run a controlled generation task.
- Test Multiple AI Models: Discover text models, generate sample inputs, compare supported options, and retrieve task results.
- Automate Cloud AI Jobs: Submit long-running app tasks without waiting, then check status and save completed results.

## Prompt Templates

### Find an App

```
Search the inference.sh store for an app that can [task]. Show the best three options and explain their documented differences. Do not run anything.
```

### Prepare a Sample Input

```
Inspect [app-name], describe its required inputs, and generate a sample input file. Ask before using any local file.
```

### Run and Track a Task

```
Run [app-name] using [input-file] without waiting. Report the task identifier, check its status, and summarize the final result.
```

### Compare a Controlled Model Workflow

```
Compare [app-one] and [app-two] for [goal]. Inspect both schemas, prepare equivalent inputs, estimate external effects, and request approval before each execution.
```

## Limitations

- Requires the belt CLI, an inference.sh account, authentication, network access, and sufficient credits.
- Available apps, models, pricing, schemas, and output quality can change outside this skill.
- Local file paths are uploaded to a cloud service and may contain sensitive data.
- Social commands can affect live accounts and require explicit user authorization.

## Best Practices

- Inspect app details and generate a sample before each first run.
- Pin app versions when repeatable output or stable schemas matter.
- Confirm costs, local file uploads, and external account actions before execution.

## Anti Patterns

- Do not pipe remote installer content directly into a shell.
- Do not upload local files until the user confirms the exact path and destination.
- Do not publish posts, messages, follows, likes, or reposts without immediate approval.

## Security Audit

- Audited at: 2026-07-12T10:06:03.275\+00:00
- Summary: Most static alerts are false positives caused by Markdown code formatting, example paths, environment variable documentation, and expected vendor links. The pipe-to-shell installer and unquoted remote-manifest command substitution are confirmed risks. Automatic local-file uploads and social account actions also require explicit user awareness and approval.

## Stats

- Views: 0
- Downloads: 5
- Favorites: 0
- Popularity score: 0