The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-generated content (comments and posts) from the Futu community platform, which creates a potential surface for indirect prompt injection.
Ingestion points: Fetches community feed data from the https://ai-news-search.futunn.com/stock_feed endpoint (identified in SKILL.md).
Boundary markers: The instructions do not specify explicit delimiters (e.g., XML tags or triple quotes) for the analysis of individual posts, although the structured workflow naturally separates data processing from aggregation.
Capability inventory: The skill is scoped to information processing, sentiment classification, and summary generation. It does not perform file-system modifications or arbitrary network calls based on data content.
Sanitization: The skill mandates cleaning steps including stripping HTML tags from titles and descriptions, filtering low-quality or spammy content, and removing near-empty text before analysis.
[COMMAND_EXECUTION]: The skill metadata specifies requirements for system binaries (curl, openssl, date). These tools are used for the primary function of retrieving data from the API and processing timestamps.
[EXTERNAL_DOWNLOADS]: The skill connects to ai-news-search.futunn.com to retrieve stock discussion data. This is a legitimate resource belonging to the service provider (Futu) and is necessary for the skill's stated purpose.

futu-comment-sentiment