The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill interacts with the redfox.hk API to retrieve video metadata and resolve direct download links. It also downloads media files from various platform CDNs to a local directory. These actions are standard for the skill's purpose as a video downloader.
[COMMAND_EXECUTION]: The skill consists of a set of Python scripts intended to be run by the agent or user via a CLI interface (main.py) for processing account-level video data.
[CREDENTIALS_UNSAFE]: Authentication is handled securely via the REDFOX_API_KEY environment variable, avoiding the risk of hardcoded secrets in the skill's codebase.
[PROMPT_INJECTION]: The SKILL.md file contains formatting requirements for the agent's output to ensure consistency. These are architectural constraints for the tool's UX and do not represent a bypass of safety protocols.
[PROMPT_INJECTION]: The skill handles untrusted external data (video titles and descriptions) obtained from third-party platforms, representing an indirect prompt injection surface. Evidence chain: 1. Ingestion points: Data is fetched via API calls in platform-specific modules (bilibili.py, douyin.py, kuaishou.py, youtube.py). 2. Boundary markers: Explicit boundary instructions for the agent are absent, though data is contained within Markdown tables. 3. Capability inventory: Python script execution, local file system writing for downloads, and outbound network requests to API endpoints. 4. Sanitization: The code in scripts/base.py performs basic character replacement (escaping pipes and removing newlines) to prevent Markdown formatting issues.

account-video-downloader