discovery-skills
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external academic APIs and RSS feeds that are then processed by the agent, creating a surface for indirect prompt injection attacks. \n
- Ingestion points: Research abstracts and titles enter the agent context via the OpenAlex API (semantic-scholar-recs-guide/SKILL.md), arXiv RSS feeds (rss-paper-feeds/SKILL.md), and the Semantic Scholar API (literature-mapping-guide/SKILL.md). \n
- Boundary markers: The skill does not implement specific delimiters or instructions to ignore instructions embedded within the research content. \n
- Capability inventory: The skill utilizes network operations to fetch data and local file-system operations to manage a vector search index (semantic-paper-radar/SKILL.md). \n
- Sanitization: There is no evidence of content sanitization or validation performed on external data before it is presented to the agent. \n- [EXTERNAL_DOWNLOADS]: The skill references and downloads external content from community and vendor repositories. \n
- Evidence: The skill includes instructions to clone the 'Papers We Love' repository (papers-we-love-guide/SKILL.md) and references a Zotero plugin provided by the vendor (zotero-arxiv-daily-guide/SKILL.md). These are legitimate, purpose-aligned resources.
Audit Metadata