The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references a remote installation script/guide located at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md to set up the necessary CLI tool.
[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute belt CLI commands. These commands are used for authentication (belt login), listing services (belt app list), and running remote applications to convert HTML to images or generate illustrations.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) because it takes user-supplied or agent-generated content and interpolates it directly into JSON payloads for remote processing.
Ingestion points: Data enters via the --input JSON string in belt app run commands, specifically within the html and prompt fields.
Boundary markers: There are no boundary markers or instructions to the agent to ignore potentially malicious content within the slide text.
Capability inventory: The skill has the capability to execute shell commands via Bash and interact with remote APIs via the belt CLI.
Sanitization: No sanitization or validation of the input strings is performed before they are passed to the shell command.

social-media-carousel