The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill references installation scripts for the Pixi package manager hosted at pixi.sh. Evidence: The commands curl -fsSL https://pixi.sh/install.sh | sh and iwr -useb https://pixi.sh/install.ps1 | iex are provided in the documentation. Context: The skill includes a specific Stop condition ('Don't run the bootstrap installer yourself') that forbids the AI agent from executing these commands automatically, ensuring they are only surfaced for manual user action.
[COMMAND_EXECUTION]: The skill is designed to execute environment management commands using standard tools including pixi, uv, poetry, hatch, conda, mamba, and pip. Evidence: The skill defines logic for running commands such as pixi add, uv sync, and pip install to manage project dependencies.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading the Pixi installer and Python packages from official and well-known sources. Evidence: References to pixi.sh and standard package registries are used as part of the intended environment management functionality.
[PROMPT_INJECTION]: The skill possesses an indirect injection surface as it reads and processes project manifest files which are externally controlled. Ingestion points: Reading pixi.toml, pyproject.toml, and requirements.txt from the project root. Boundary markers: Implements a mandatory 'Pre-flight' checklist and structured 'Stop conditions' to validate tool choices and scopes. Capability inventory: Ability to execute shell commands via package manager CLI tools. Sanitization: Actions are gated by structured user confirmation requests (AskUserQuestion).

python-env-manager