The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the skore_skills CLI tool for administrative tasks such as workspace scaffolding (scaffold), policy management (policy), and session control (git end-turn). These commands are part of the vendor's own toolset for project management.
[EXTERNAL_DOWNLOADS]: The skill triggers the installation of several well-known and legitimate data science packages, including jupytext, nbclient, ipywidgets, nbconvert, and mkdocs-material. These installations are performed through the add-python-package skill and represent standard requirements for the described machine learning workspace.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it reads local project data to determine layout state.
Ingestion points: Layout detection logic reads the presence of directories (src/, journal/) and the content of pyproject.toml as described in SKILL.md.
Boundary markers: There are no explicit markers or instructions to ignore instructions embedded in the project files.
Capability inventory: The skill can create project directories and files via skore_skills scaffold, update workspace settings via skore_skills policy, and initiate package installs as defined in SKILL.md.
Sanitization: The skill follows a strict lookup table and restricts package names to a predefined list of reputable tools, mitigating the risk of arbitrary command execution from malicious project metadata.

setup-workspace