The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the agent to generate and run Python scripts at runtime to perform statistical calculations and data processing based on user-provided genomic data. The ANALYSIS_PROCEDURES.md file provides extensive logic templates and pseudocode that the agent uses to construct these scripts which incorporate data from both the user and external tools.
[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection as it ingests untrusted user data, such as genomic coordinates and clinical phenotypes, which are then processed using high-privilege tools and dynamic code execution. The analysis identified the following evidence chain: 1. Ingestion points: User-supplied genomic data and phenotypes in the initial query. 2. Boundary markers: Absent in the workflow description. 3. Capability inventory: Bash tool for Python execution and multiple biological data retrieval tools. 4. Sanitization: No specific sanitization or filtering logic is defined for the interpolated user input.
[COMMAND_EXECUTION]: The agent is explicitly instructed to use a bash tool to execute the Python scripts it generates, which provides a mechanism for running commands within the agent's execution environment.
[EXTERNAL_DOWNLOADS]: The skill fetches scientific evidence and genomic data from well-known external services including ClinVar, gnomAD, and PubMed. These are legitimate resources used for clinical interpretation purposes.

tooluniverse-structural-variant-analysis