The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions describe building and executing ffmpeg shell commands dynamically using parameters provided in the input configuration, such as clips, output, bgm, and subtitles.- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect injection because it ingests untrusted external data that influences shell command generation.
Ingestion points: Parameters including clips (URLs/paths), bgm, subtitles, and output specified in SKILL.md.
Boundary markers: Absent. No delimiters or instructions are provided to the agent to treat these inputs as untrusted or to ignore embedded control characters.
Capability inventory: Execution of ffmpeg via shell commands, which has the capability to read any file the user has access to and write to specified locations.
Sanitization: Absent. There is no mention of validating, escaping, or sanitizing the input paths before they are interpolated into shell commands.- [REMOTE_CODE_EXECUTION]: The use of the -safe 0 flag in the ffmpeg -f concat command allows the tool to process files from any path. If the clips list is sourced from an untrusted upstream skill or user input containing shell metacharacters, it could lead to arbitrary command execution.

video-stitcher