The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install a Python wheel (tuplatform_connect) from a remote URL (https://connect.aiscientist.tools/downloads/...). The domain aiscientist.tools is an external source for executable code packages, which may lead to the installation of unverified software.
[REMOTE_CODE_EXECUTION]: By instructing the installation of an external package via a direct URL (pip install [URL]), the skill facilitates the execution of code contained within that package during installation or subsequent use. While a SHA-256 hash is provided, the package originates from an unverified third-party domain.
[INDIRECT_PROMPT_INJECTION]: The run_squidpy_nhood_enrichment operation accepts an adata_path file as input, creating a data ingestion surface where maliciously crafted files could influence agent behavior. The skill lacks explicit sanitization or boundary markers for these inputs.
Ingestion points: The adata_path parameter in the run_squidpy_nhood_enrichment tool.
Boundary markers: No explicit delimiters are mentioned to isolate untrusted file content.
Capability inventory: The skill calls MCP tools, interacts with the local file system through Python scripts, and manages remote tool connections.
Sanitization: No sanitization or validation of the input file content is documented.
[COMMAND_EXECUTION]: The skill relies on multiple shell commands for environment setup, CLI authentication, and tool deployment (tu remote login, tu remote share, pip install).

setup-squidpy-remote-tool