The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill demonstrates workflows where the output of one AI model (e.g., a generated video script) is used as the direct input for another model (e.g., a text-to-speech generator). This pattern allows for potential injection if a generated script contains instructions that the downstream model might mistakenly follow.
Ingestion points: The 'Explainer Video' section captures the output of the script generation tool and passes the raw content to the voiceover generation tool.
Boundary markers: The skill does not employ delimiters or system instructions to distinguish between the data being processed and the instructions for the model.
Capability inventory: The skill utilizes the Bash(belt *) tool, providing it with the capability to execute complex CLI operations, perform network requests through the platform API, and write files to the local system.
Sanitization: The instruction set lacks mechanisms to sanitize or validate the generated text before it is interpolated into subsequent tool prompts.
[EXTERNAL_DOWNLOADS]: The skill references external installation documentation located at raw.githubusercontent.com/inference-sh/skills and suggests adding the belt-sh/cli package. It also illustrates media merging using a placeholder external URL (https://trending-music.mp3).
[COMMAND_EXECUTION]: The skill relies on executing belt CLI commands through the Bash tool to perform model inference, handle JSON outputs, and manage media assembly workflows.

ai-marketing-videos