The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions include downloading a Python wheel from a non-standard external domain.
Evidence: pip install ... "tuplatform-connect @ https://connect.aiscientist.tools/downloads/tuplatform_connect-0.3.0-py3-none-any.whl#sha256=3fad5eee5ecf7887a693d93ccd1aa112dc0955617a885d1fc3daded0030f9ae0" in SKILL.md.
Note: The download is pinned with a SHA-256 hash, which provides integrity protection.
[COMMAND_EXECUTION]: The skill requires the execution of multiple shell commands to set up virtual environments, install packages, and manage remote connections via the tu CLI.
Evidence: Commands like python scripts/remote_validation/setup_skill_preflight.py, tu remote login, and tu serve are required for the skill to function.
[INDIRECT_PROMPT_INJECTION]: The skill processes biological data in .h5ad format, which creates a surface for indirect prompt injection if the data files contain malicious metadata or instructions.
Ingestion points: The adata_path parameter in the run_cellrank_fate operation (SKILL.md).
Boundary markers: None identified in the prompt interpolation logic.
Capability inventory: The skill spawns a local server and executes computational kernels using the cellrank library.
Sanitization: Instructions mention checking for scientific meaning and finite values, but do not specify automated sanitization of the input file content.

setup-cellrank-remote-tool