scrna-celltype-annotation
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's primary function is to perform biological data analysis and literature cross-referencing, which is a legitimate and safe use case for an AI agent.\n- [EXTERNAL_DOWNLOADS]: The skill uses MCP tools to interface with PubMed and OpenAlex. These are well-known academic and medical services. The references are used to fetch literature abstracts and metadata to support cell type annotation, posing no security risk.\n- [COMMAND_EXECUTION]: While the skill generates R code for 'recode' operations, it does so as text output for the user's manual use. There is no evidence of the skill attempting to execute these or any other system commands automatically.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from user-provided CSV files (Ingestion point: SKILL.md, Workflow Step 2). Although it lacks explicit boundary markers for individual CSV fields, its primary capability is writing Markdown reports and R code snippets (Capability inventory) based on this data. Since it does not perform automated shell execution or network exfiltration of sensitive files, the injection surface is considered low risk. No specific sanitization of CSV content is described.
Audit Metadata