The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically generates a JSON configuration file (showcase-props.json) at a temporary location. The contents of this file are later passed as an argument to the npx remotion command. This pattern of generating configuration data that influences the execution of a rendering engine represents a dynamic execution surface.
[COMMAND_EXECUTION]: The skill relies on executing various shell commands and scripts, including render-showcase.sh, npx remotion, ffmpeg, ffprobe, agg, and asciinema. These commands are constructed using parameters (such as file paths, labels, and title text) derived from external inputs.
[INDIRECT_PROMPT_INJECTION]: The skill explicitly ingests "Creative (natural language)" guidance to drive editorial decisions. This untrusted data represents an attack surface for indirect prompt injection if it influences the generation of shell commands or the JSON properties without adequate sanitization.
Ingestion points: The skill processes both mechanical structured data and natural language guidance defined in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the creative guidance as potentially untrusted or to ignore embedded instructions.
Capability inventory: The skill has the ability to write files (cat), execute shell scripts (render-showcase.sh), run Node.js tools (npx remotion), and perform video/terminal processing (ffmpeg, agg, asciinema).
Sanitization: The documentation does not specify any sanitization, escaping, or validation steps for the natural language input before it is used to build the execution properties.

compose