The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references installation instructions for the belt CLI hosted on GitHub at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
[COMMAND_EXECUTION]: The skill extensively uses the belt CLI tool (authorized in the allowed-tools frontmatter) to perform various AI-driven tasks via the Bash shell:
Authenticating with the inference service using belt login.
Executing various remote applications for image generation (falai/flux-dev-lora), text-to-speech (falai/dia-tts), video generation (google/veo-3-1-fast), and social media automation (x/post-create).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user-provided data (e.g., blog posts, podcast transcripts) and interpolate it into templates for downstream AI tool calls.
Ingestion points: Untrusted long-form content is ingested and processed in SKILL.md via the 'Conversion Recipes' section.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious instructions embedded in the source content.
Capability inventory: The skill utilizes Bash to execute the belt tool, which has capabilities including media generation and social media posting.
Sanitization: The instructions do not define sanitization or escaping protocols for the content being repurposed before it is passed to the AI applications.

content-repurposing