The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of local shell commands via bash scripts to manage media processing and network requests.
Evidence: All scripts in the scripts/ directory invoke utilities such as ffmpeg, ffprobe, jq, curl, and xxd to convert, merge, and download media files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text for AI generation, which presents a surface for indirect prompt injection attacks.
Ingestion points: Input arguments like --prompt, --lyrics, and --scenes in scripts/image/generate_image.sh, scripts/music/generate_music.sh, scripts/video/generate_video.sh, and scripts/video/generate_long_video.sh.
Boundary markers: None identified. The instructions do not define delimiters to wrap user content or specify instructions to ignore embedded commands when processing external text.
Capability inventory: The skill has the ability to execute shell commands, perform network operations via curl, and write files to the filesystem.
Sanitization: Payloads for API requests are constructed using jq with the --arg flag, which protects against shell injection into the JSON structure but does not filter the semantic content of the prompts themselves.

minimax-multimodal-toolkit