evolutionary-biologist
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a comprehensive workflow for ingesting and analyzing external scientific data, creating an attack surface where an attacker could embed malicious instructions within genomic data files or their associated metadata. This is a characteristic of indirect prompt injection vulnerability surfaces.
- Ingestion points: The workflow instructs the agent to handle raw sequencing reads (FastQ), alignments (BWA-MEM, minimap2), and variant calls (VCF files) in SKILL.md.
- Boundary markers: The skill lacks explicit instructions or delimiters for the agent to treat content within these scientific files as untrusted data or to ignore embedded natural language instructions.
- Capability inventory: The skill provides the agent with a list of powerful command-line tools for execution, including BWA-MEM, GATK, bcftools, IQ-TREE, and RAxML-NG, which are intended to process the external data.
- Sanitization: While the skill recommends biological quality control (e.g., filtering for minor allele frequency or depth), it does not include security-focused sanitization or validation of the data content to prevent instruction injection.
Audit Metadata