xhs-explore
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content originating from the Xiaohongshu platform via the
list_feedsandget_feed_detailtools. - Ingestion points: Notes, titles, descriptions, and user comments retrieved from external API responses (SKILL.md).
- Boundary markers: The instructions do not include specific delimiters or directives to the agent to treat fetched social media text as data only, potentially allowing embedded instructions in posts to influence agent behavior.
- Capability inventory: While this skill focuses on retrieval, it directs the agent toward other capabilities such as interacting with posts (
xhs-interact) or viewing profiles (xhs-profile) based on the data ingested. - Sanitization: There is no evidence of content filtering, escaping, or validation to mitigate malicious instructions within the social media data stream.
Audit Metadata