The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches content from the external Hacker News Algolia API, which contains unmoderated, user-generated text (titles, story text, and comments).
Ingestion points: The search_hn function in scripts/search_hn.py retrieves hits from hn.algolia.com and includes fields like title, story_text, and comment_text in the agent's context.
Boundary markers: There are no specific boundary markers or instructions to the agent to ignore potentially malicious instructions embedded within the fetched HN content.
Capability inventory: The skill uses requests for network access to the Algolia API and prints results to stdout for the agent to process. The underlying environment may have additional tools accessible to the agent.
Sanitization: No sanitization or filtering is performed on the text or title fields before they are output and read by the LLM, creating a surface for prompt injection attacks hidden in HN posts or comments.

hacker-news-scraper