The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the ffmpeg and ffprobe utilities for video merging and duration detection. These commands are executed via Python's subprocess.run with arguments constructed from internally managed task structures, minimizing the risk of arbitrary command injection.
[EXTERNAL_DOWNLOADS]: Instructions are provided to install ffmpeg via standard package managers (brew, apt, yum). ffmpeg is a well-known, trusted, and widely used open-source tool for multimedia processing.
[PROMPT_INJECTION]: The skill includes a 'Content Safety Audit' section that instructs the agent to identify and rephrase sensitive content (e.g., replacing 'blood' with 'spiritual energy') before submitting prompts to the generation API. This is a functional requirement to prevent task failure due to external API content filters and does not constitute a malicious attempt to bypass the agent's own safety guardrails.
[DATA_EXFILTRATION]: Network operations are directed towards official Volcengine domains (*.volces.com and *.volcengineapi.com) for image/video generation, search, and storage. These actions are transparently documented and align with the primary purpose of the skill and the author's identity (Bytedance).

comic-drama-master