The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the R package 'CellChat' directly from a GitHub repository ('jinworks/CellChat') using 'devtools::install_github'. It also recommends installing the 'liana' package via 'pip'. These actions fetch code from third-party sources outside of the primary trusted registries.- [REMOTE_CODE_EXECUTION]: The process of installing packages from source repositories like GitHub via 'devtools' involves the compilation and execution of code provided by the repository author on the user's system.- [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external data. * Ingestion points: The skill processes user-supplied scRNA-seq count matrices and metadata objects ('seurat_obj', 'data.input', 'meta'). * Boundary markers: The instructions do not define boundary markers or provide 'ignore embedded instructions' warnings for the data being processed. * Capability inventory: The skill has capabilities to write files ('write.csv', 'ggsave'), read files ('readRDS'), and generate visualizations. * Sanitization: There is no explicit evidence of sanitization or validation of input data fields (such as cell type labels or cluster names) before they are interpolated into plot labels or exported to tables.

cellchat-cell-communication