Quick Install
Install the Sher skill for your AI agent:sher-deploy skill, which teaches agents how to use the Sher CLI to deploy frontend projects and return live preview URLs.
What the Skill Does
The Sher skill provides your AI agent with:- Knowledge of Sher commands — how to run
sher linkand when to use deployment flags - Framework detection — understanding of which frameworks Sher supports
- Output parsing — how to extract preview URLs from CLI output
- Error handling — troubleshooting common deployment issues
- Best practices — when to deploy, how to present URLs to users
Prerequisites
Before installing the skill, ensure:Installation Steps
Run the install command
Execute the skill installation command:This downloads and registers the Sher deployment skill with your agent.
Verify installation
The skill is now available to your agent. No additional configuration is needed.
How Agents Use the Skill
Once installed, your agent can deploy projects automatically. Here’s the typical workflow:Example Agent Workflow
Deployment Options
The skill teaches agents to use these flags appropriately:Deploy a specific directory (skips framework detection)
Set link expiration in hours
Skip the build step if project is already built
Password-protect the preview (Pro tier only)
Parsing URLs from Output
The skill teaches agents to extract preview URLs using this pattern:sher.sh.
Presenting URLs to Users
The skill includes best practices for presenting deployed URLs:Troubleshooting
Skill installation fails
Skill installation fails
Ensure Node.js and npm are installed and accessible in the agent’s environment. The
skills CLI requires npm to download and install skill packages.Agent can't find sher command
Agent can't find sher command
Install Sher globally in the agent environment:Alternatively, the agent can use
npx shersh link instead of sher link.Deployment fails with size limit error
Deployment fails with size limit error
The project exceeds the upload size limit:
- Anonymous: 10 MB max
- Authenticated: 50 MB max
- Pro: 100 MB max
No output directory found
No output directory found
The build didn’t produce output in the expected location. The agent should:
- Check that a
buildscript exists inpackage.json - Verify the build completed successfully
- Use
--dirto specify the output directory explicitly if needed
Configuration
No additional configuration is required after skill installation. The skill is self-contained and works immediately.Optional: Authentication
For better rate limits (25 links/day instead of 1), agents can authenticate with GitHub:Supported Frameworks
The skill includes knowledge of all supported frameworks:Vite
Auto-detects Vite projects and builds to
dist/Next.js
Configures static export automatically, builds to
out/Astro
Detects Astro projects and builds to
dist/Create React App
Builds CRA projects to
build/build script that produces dist/, build/, or out/ directories.
Next Steps
Link Command
Learn about deployment command options and flags
Agent Overview
Understand AI agent integration capabilities
List Command
View and manage active deployments
Authentication
Set up GitHub authentication for better limits