The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and fetches installation instructions from an external GitHub repository (https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md). It also instructs the user to install a CLI package (belt-sh/cli) using npx.
[COMMAND_EXECUTION]: The skill utilizes the belt command-line tool via Bash to perform operations such as logging into the service and executing remote AI models. This is configured in the frontmatter's allowed-tools field.
[INDIRECT_PROMPT_INJECTION]: The skill provides templates for generating images and videos where user-supplied or predefined prompts are interpolated into shell commands. This creates a surface for indirect prompt injection if untrusted data is processed through these templates.
Ingestion points: Prompt fields within the JSON payload of belt app run commands in SKILL.md.
Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands within the prompt strings.
Capability inventory: The skill can execute the belt CLI tool which interacts with remote inference APIs.
Sanitization: Absent; the skill does not show any escaping or validation logic for the text passed to the models.

app-store-screenshots