The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads a binary Python wheel file (tuplatform_connect-0.3.0-py3-none-any.whl) from the external domain connect.aiscientist.tools. This domain is not a recognized trusted vendor or a standard package registry.
[REMOTE_CODE_EXECUTION]: The skill instructs the agent to use pip install to install software directly from a remote URL (https://connect.aiscientist.tools/...). Installing binary wheels from unverified sources allows for the execution of arbitrary code during the installation process and the inclusion of potentially malicious modules in the local environment.
[COMMAND_EXECUTION]: The skill requires the execution of multiple high-risk shell commands, including manual virtual environment management, the installation of local requirements files, and the execution of dynamically generated Python scripts using heredocs (`python
<<'PY'`).
[PROMPT_INJECTION]: The skill handles external data that creates a surface for indirect prompt injection attacks.
Ingestion points: The skill reads provider artifacts from paths specified in the PINNACLE_DATA_PATH environment variable and processes user-provided JSON arguments for the retrieval tool.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the processing logic described.
Capability inventory: The environment allows package installation, local file system access, and the creation of network relays using the tu serve command.
Sanitization: There is no evidence of input validation or sanitization for data retrieved from external artifacts or user arguments.

setup-pinnacle-remote-tool