The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install a Python wheel ('tuplatform-connect') directly from a remote URL ('https://connect.aiscientist.tools/downloads/...') rather than a standard package registry like PyPI. Installing software from non-standard sources introduces supply chain risk.
Evidence: python -m pip install fastmcp pyyaml "tuplatform-connect @ https://connect.aiscientist.tools/downloads/tuplatform_connect-0.3.0-py3-none-any.whl#sha256=3fad5eee5ecf7887a693d93ccd1aa112dc0955617a885d1fc3daded0030f9ae0" in SKILL.md.
[COMMAND_EXECUTION]: The skill requires the user to execute numerous shell commands, including environment creation, credential management, and starting local HTTP/MCP servers.
Evidence: Multiple code blocks in SKILL.md using pip, venv, and a custom CLI tool (tu).
[INDIRECT_PROMPT_INJECTION]: The run_enformer_predict tool accepts external genomic sequences as input, which serves as a potential ingestion point for malicious instructions meant to influence the model's behavior.
Ingestion points: The sequence argument in the run_enformer_predict tool implementation described in SKILL.md.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands are present in the documentation for sequence processing.
Capability inventory: The skill environment includes capabilities to execute shell commands via the tu tool and host local servers.
Sanitization: There is no evidence of sequence sanitization or validation beyond checking for scientific meaning.

setup-enformer-remote-tool