The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch installation instructions from a remote GitHub repository at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to use the belt CLI and npx to install tools and manage logins, which involves executing system-level commands.
[REMOTE_CODE_EXECUTION]: The skill uses belt app run to trigger the execution of remote services (such as html-to-image and flux-dev-lora). The actual logic for these tools is executed on external infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill constructs HTML strings using user-provided titles and descriptions. This creates a potential surface where malicious input could be rendered by the remote service, although the risk is limited by the context of image generation.
Ingestion points: Title, subtitle, and branding text provided by the user for the OG image.
Boundary markers: None are specified; the user input is directly concatenated into the HTML style and content strings.
Capability inventory: Remote image rendering and search assistance via the belt toolset.
Sanitization: The skill does not provide instructions for escaping or sanitizing the input before it is processed by the renderer.
[DYNAMIC_EXECUTION]: The skill generates HTML code blocks at runtime based on templates and then passes these strings to a remote execution tool for rendering.

og-image-design