The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data from Twitter/X, including tweets, user bios, and comments. This data could contain malicious instructions aimed at influencing the AI agent's subsequent logic.
Ingestion points: External data enters the agent context via multiple tools defined in references/sandbase-api-map.md, such as twitter_web_search_timeline, twitter_web_user_profile, and twitter_web_tweet_detail.
Boundary markers: The instructions in SKILL.md lack explicit delimiters or warnings to treat the retrieved content as data rather than instructions.
Capability inventory: The skill uses sandbase_call_tool to fetch external content. If used in a broader environment with write capabilities, the content of a tweet could potentially trigger unauthorized actions.
Sanitization: No specific sanitization or escaping protocols are provided for handling retrieved tweet content before analysis.

twitter-intelligence