The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the RedFox API, specifically video descriptions (aweme_desc) and author names (user_nickname).
Ingestion points: The call_api function in scripts/douyin_weekly_surge.py retrieves content data which is then rendered in Markdown tables for the user.
Boundary markers: No explicit instructions or delimiters are used to warn the agent against potential instructions embedded in the video descriptions.
Capability inventory: The skill has limited high-risk capabilities; it primarily outputs data to the terminal and uses a specific CLI tool for payments. It does not perform arbitrary file writes or general network operations based on the fetched data.
Sanitization: The script scripts/douyin_weekly_surge.py performs basic sanitization by replacing Markdown-sensitive characters (like |, [, ]) in video titles to prevent layout issues.
[EXTERNAL_DOWNLOADS]: The skill documentation in references/alipay-recharge.md instructs the agent to install an external package from a well-known service if it is missing.
Evidence: Execution of npx -y @alipay/agent-payment@latest install to provide payment capabilities.
[COMMAND_EXECUTION]: The skill utilizes an external CLI tool to manage automated payment flows.
Evidence: Usage of the alipay-bot command-line utility with subcommands like trigger-payment-signal, submit-payment, and poll-payment-status to handle API credit recharging.
[DATA_EXFILTRATION]: The skill communicates with the vendor's API endpoint at redfox.hk.
Evidence: The script scripts/douyin_weekly_surge.py sends POST requests to https://redfox.hk/story/api/dy/search/hotContentRank using an API key provided in environment variables. This is a standard vendor resource and no sensitive user data exfiltration was detected.

douyin-weekly-surge