# Run inference.sh AI Apps

Agents need one interface for many hosted AI models and media workflows. This skill teaches Claude, Codex, and Claude Code to use the infsh CLI safely.

## Install

```bash
npx skillstore add inferencesh/agent-tools
```

## Metadata

- Status: approved
- Slug: inferencesh-agent-tools
- Skillstore revision: r1
- Version status: missing
- Tree hash: 1a0e297d828386e79981efc26642de2407522b1c6acab4be86b8845ca92d6da3
- Author: inferencesh
- GitHub username: inferencesh
- License: MIT
- Repository: https://github.com/inferencesh/skills/tree/main/skills/agent-tools/
- Ref: a25199bc7d6b82598536822d1738eb5d5f54025b
- 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/inferencesh-agent-tools
- Manifest: https://skillstore.pages.dev/api/skills/inferencesh-agent-tools/manifest

## Capabilities

- Install, update, and authenticate the infsh CLI.
- List, search, filter, and inspect inference.sh apps.
- Run AI apps with inline input or input files.
- Track long-running tasks and save task results.
- Use examples for images, videos, LLM calls, search, 3D, audio, and X/Twitter actions.

## Use Cases

- Prototype Media Generation: Find image, video, audio, or 3D apps and run fast test jobs from the command line.
- Compare Hosted AI Models: Discover LLM, search, and media apps, inspect schemas, and compare results across providers.
- Operate Account Workflows: Use authenticated CLI commands for app runs, task tracking, deployment checks, and optional X/Twitter workflows.

## Prompt Templates

### List Available Apps

```
Find inference.sh apps for image generation. List strong options and explain what each one is best for before running anything.
```

### Create Sample Input

```
Inspect the app google/veo-3-1-fast and create a sample input file. Explain which fields I should edit.
```

### Run And Track A Task

```
Run an image generation task with my prompt, track it until completion, and summarize the returned media result.
```

### Build A Multi-Step Workflow

```
Find a text-to-image app, prepare a prompt from my product brief, run the task, then collect supporting references with a search app. Ask before paid or publishing actions.
```

## Limitations

- Requires an inference.sh account and local infsh CLI access.
- Cloud apps may require credits, provider permissions, or third-party credentials.
- Generated media, search results, and model outputs depend on the selected app.
- Installation guidance includes remote shell installer commands that require manual trust review.

## Best Practices

- Review remote installation commands before execution.
- Use sample input and app details before running paid tasks.
- Confirm X/Twitter, deployment, or credential-related actions with the account owner.

## Anti Patterns

- Do not pipe remote installers to a shell without verification.
- Do not run publishing or account actions without explicit approval.
- Do not place real API keys in prompts, logs, or shared files.

## Security Audit

- Audited at: 2026-07-05T17:37:52.47\+00:00
- Summary: Review confirmed multiple remote installer commands that pipe downloaded content directly into a shell, which is a critical supply-chain risk. Most API key, hidden file, documentation link, and system reconnaissance matches are benign reference material. The skill still carries elevated risk because it grants broad infsh command access and includes account-changing cloud and X/Twitter actions.

## Stats

- Views: 201
- Downloads: 12
- Favorites: 0
- Popularity score: 0