The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references external installation instructions and remote files hosted on GitHub.
Evidence: https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md referenced in SKILL.md.
[COMMAND_EXECUTION]: The skill instructs the user to install third-party CLI tools and agent extensions from external repositories.
Evidence: npx skills add belt-sh/cli and npx skills add inference-sh/skills@... commands in SKILL.md.
[COMMAND_EXECUTION]: The skill utilizes the belt CLI to execute remote applications, facilitating interactions with external APIs and services.
Evidence: Multiple instances of belt app run targeting remote models like falai/flux-dev-lora and tavily/search-assistant.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a potential attack surface for indirect prompt injection by interpolating user-provided text into shell commands for remote execution.
Ingestion points: Prompt fields within the --input parameters of belt app run commands.
Boundary markers: None; external inputs are directly included in JSON command payloads.
Capability inventory: The skill is configured to allow Bash(belt *) execution.
Sanitization: No input validation or escaping mechanisms are described for the processed text.

landing-page-design