The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of Douyin video titles and descriptions obtained from the RedFox API. This content is used by the agent to generate data highlights and account feature analysis, representing a surface for indirect prompt injection. However, the risk is inherent to the skill's primary purpose and the data is handled through structured API responses. 1. Ingestion points: Account metrics and work data containing titles fetched in scripts/douyin_works_fetcher.py. 2. Boundary markers: The output template in references/core_workflow.md uses markdown tables and blockquotes to organize external content. 3. Capability inventory: The skill performs network requests to the vendor domain redfox.hk and executes local Python scripts. 4. Sanitization: Video titles are truncated in the output to mitigate potential injection impact.

douyin-works-crawler