The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses ffmpeg for media assembly and processing. The assemble_mv.sh script constructs ffmpeg commands using an array-based approach and implements helper functions like safe_concat_entry to escape filenames and paths, mitigating potential command injection. The test suite (tests/) utilizes os.system and subprocess.run to generate synthetic media and execute the bash scripts for integration testing, which is standard practice for CLI automation tools.\n- [EXTERNAL_DOWNLOADS]: The scripts suno_music.sh and gen_visuals.sh perform network requests to external AI service providers (SunoAPI, OpenAI, BytePlus, and Together AI) to generate and download music, images, and videos. These operations are core to the skill's functionality and target well-known service endpoints. The suno_music.sh script also provides a configurable callback URL, which defaults to a safe local no-op endpoint and includes validation to ensure only HTTPS schemes are used.\n- [DYNAMIC_EXECUTION]: The bash scripts extensively use python3 -c to parse and generate JSON data. This is used for constructing API request bodies and tracking token-based costs. Payloads are constructed using json.dumps, ensuring that user-provided prompts and metadata are properly escaped before being sent to external APIs or used in shell environments.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts for music and visual generation ingested via SKILL.md or prompts.json. While these prompts are interpolated into API requests without specific boundary markers, the skill uses structured JSON serialization (json.dumps) for payload construction and implements path escaping for its ffmpeg assembly capabilities. Capabilities are limited to media generation and local file operations within the specified output directory, with sanitization provided by robust JSON handling in the automation scripts.

ai-music-video