Skills agent-tools
🤖

agent-tools

Medium Risk ⚙️ External commands🌐 Network access🔑 Env variables📁 Filesystem access

Run 250+ AI Models via inference.sh CLI

Also available from: qu-skills,inference-sh-8,inference-sh-3,inference-sh-6,toolshell,inferen-sh,inf-sh,infsh-skills,inference-sh-7,inferencesh,inference-shell,inference-sh-skills,tul-sh,inference-sh-0,inference-sh-9,skillssh,tool-belt

Need to generate images, create videos, or call AI models? This skill wraps the inference.sh CLI to give Claude access to 250+ AI services including FLUX, Veo, Gemini, and Claude itself.

Supports: Claude Codex Code(CC)
⚠️ 65 Poor
1

Download the skill ZIP

2

Upload in Claude

Go to Settings → Capabilities → Skills → Upload skill

3

Toggle on and start using

Test it

Using "agent-tools". Generate a landscape image of a mountain at sunset using FLUX

Expected outcome:

Image generation initiated via inference.sh... Download URL provided upon completion

Using "agent-tools". Call Gemini to summarize: The quick brown fox jumps over the lazy dog

Expected outcome:

Summary: A concise description of a fox's agility demonstrated through a common pangram

Security Audit

Medium Risk
v1 • 5/4/2026

Static analysis flagged 203 potential issues, primarily pattern detection. After semantic evaluation: backtick/shell commands are the legitimate CLI wrapper mechanism, not injection vulnerabilities. Hardcoded URLs are documentation examples. The skill combines code execution, network access, and credential handling to call external AI services - this is documented functionality, not malicious behavior. No evidence of data exfiltration or malicious intent. Risk factors are inherent to CLI wrapper operations.

5
Files scanned
596
Lines analyzed
7
findings
1
Total audits
Medium Risk Issues (1)
CLI Wrapper Inherently Combines Elevated-Risk Operations
The skill legitimately combines code execution, network access, and credential handling to call external AI services via inference.sh. This is the documented purpose. No malicious exfiltration or unauthorized access detected.
Low Risk Issues (2)
MD5 References in Documentation
Static analyzer flagged MD5 usage as 'weak cryptographic algorithm'. These appear in documentation examples showing how to compute hashes, not in security-critical code paths.
Hardcoded URLs in Documentation
Static analyzer flagged hardcoded URLs. These are API endpoint documentation examples, not actual runtime network calls to untrusted destinations.
Audited by: claude

Quality Score

45
Architecture
100
Maintainability
85
Content
31
Community
56
Security
91
Spec Compliance

What You Can Build

Generate Marketing Images

Create product images, social media graphics, or marketing assets using FLUX or Seedance models directly from Claude

Access Claude via External CLI

Use Claude models through the inference.sh wrapper for batch processing or scripted workflows

Research Web Search

Query web search services like Tavily or Exa to gather research information for analysis

Try These Prompts

Generate an image
Use inference.sh to generate an image of [description] using FLUX model
Create video content
Run the following video generation using inference.sh with Veo: [scene description]
Call LLM for analysis
Using inference.sh, send the following text to Gemini for sentiment analysis: [text]
Web search for research
Search the web using inference.sh with Tavily for information about [topic]

Best Practices

  • Set INFERENCE_API_KEY in environment before using skill features
  • Check rate limits before running batch operations on AI services
  • Verify output from AI models before using in production systems

Avoid

  • Do not pass untrusted user input directly as inference.sh arguments without validation
  • Do not expose API keys in skill configurations or shared prompts
  • Do not rely solely on AI-generated content without human review

Frequently Asked Questions

How do I get an inference.sh API key?
Sign up at inference.sh to obtain an API key. Set it as INFERENCE_API_KEY environment variable.
Which AI models are supported?
The skill supports 250+ models including FLUX, Veo, Gemini, Grok, Claude, OpenRouter, Tavily, and Exa.
Are there rate limits?
Rate limits depend on your inference.sh plan and the specific AI service being accessed.
Can I use this to call Claude?
Yes, you can access Claude through inference.sh. Set ANTHROPIC_API_KEY and use the Claude model option.
Is image generation free?
Image generation costs depend on the model provider. Check inference.sh pricing for details.
Can I automate Twitter?
Yes, the skill supports Twitter API integration. Configure your Twitter API credentials properly.