The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on external binaries for media processing and rendering. Several scripts (main.py, narration.py, media.py, timeline.py, and stage.py) execute shell commands via subprocess and os.system. Of note, scripts/stage_engine/stage.py uses os.system with f-string interpolation to call ffmpeg, using paths and filenames defined in the brief.json configuration. This pattern is vulnerable to command injection if an attacker-controlled configuration file is processed.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to retrieve necessary assets and data from remote services. It fetches SVG icons from cdn.jsdelivr.net, communicates with api.openai.com for narration and image generation, and retrieves brand icon data from raw.githubusercontent.com and thesvg.org. These operations are consistent with the skill's documented purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes configuration and script files (brief.json and vo.txt) that can influence its internal logic and shell command execution.
Ingestion points: Configuration parameters and narration text are read from files within the project directory.
Boundary markers: None observed; inputs are directly interpolated into templates and commands.
Capability inventory: The skill has the ability to execute shell commands, manage files, and access the network.
Sanitization: There is no evidence of input validation or escaping for the text and paths before they are used in shell commands.

kinetic-launch-film