The Agent Skills Directory

Remote Code Execution (HIGH): An automated scan detected a piped remote execution pattern: curl -fsSL https://pixi.sh/install.sh | sh. While this is a common installation method for developer tools, it executes remote code without verification and the source domain (pixi.sh) is not on the trusted external sources whitelist.
Indirect Prompt Injection (LOW): The skill is designed to process and act upon content from manifest files (pixi.toml, pyproject.toml) which may be provided by untrusted third parties in a repository context.
Ingestion points: pixi.toml, pyproject.toml (README.md, references/manifest.md).
Boundary markers: None detected; the agent is instructed to use these files as authoritative for project configuration.
Capability inventory: The skill enables subprocess execution via pixi run, pixi shell, and pixi task add (references/commands.md).
Sanitization: None detected; the skill relies on the Pixi CLI to parse and execute commands defined in the manifest.
Command Execution (LOW): The skill documentation describes extensive use of the pixi CLI to manage environments and run tasks. While these are legitimate functions of the tool, they provide a powerful interface for an agent to execute arbitrary commands if directed by a malicious manifest.

pixi