The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the Velt SDK through the vendor's official CDN and GitHub repository.
Evidence: References to https://cdn.velt.dev/lib/sdk@latest/velt.js for vanilla JS integration and github.com/anthropics/agent-skills for skill installation.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to install necessary dependencies using standard package managers.
Evidence: Instructions for npm install @veltdev/react and npm install @veltdev/client found in installation rules.
[INDIRECT_PROMPT_INJECTION]: The skill provides instructions to fetch and process data from external documentation sources, creating a potential surface for indirect injection if the remote documentation were compromised.
Ingestion points: Remote documentation URLs (https://docs.velt.dev/) referenced throughout the rule files and metadata.json.
Boundary markers: Not present; the skill assumes the fetched documentation content is authoritative.
Capability inventory: The skill has capabilities to perform network operations (fetch calls in generated code) and subprocess execution (npm install instructions).
Sanitization: Not applicable as the content is instructional; however, the skill explicitly teaches secure coding practices to prevent credential exposure.

velt-setup-best-practices