The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is configured to fetch real-time social media trend data from external third-party API providers, including 60s.viki.moe and v2.xxapi.cn. While these are used to retrieve public hotlists for platforms like TikTok (Douyin), Bilibili, and Weibo, they represent network operations to non-whitelisted domains.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted content such as user comments and search results from platforms like Xiaohongshu and Bilibili, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters the context through web_fetch calls to social media content pages and comment sections as described in SKILL.md (Step 3) and references/demand-signals.md.
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from executing instructions that might be embedded within the social media comments it processes.
Capability inventory: The skill utilizes web_search and web_fetch capabilities; however, no file system write operations or arbitrary shell command execution were identified.
Sanitization: The instructions do not specify any sanitization, filtering, or escaping mechanisms for the external data before it is processed by the agent.

skill-content-gap-analysis