The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions to append configuration lines to shell profile files, including ~/.bashrc, ~/.zshrc, and ~/.config/fish/config.fish. This is a standard persistence mechanism for shell utilities to ensure the tool is available in every new session.
[REMOTE_CODE_EXECUTION]: The documentation demonstrates how to add plugins using custom Git URLs (e.g., https://github.com/org/asdf-my-tool.git). Because asdf plugins consist of shell scripts that are executed on the host system to manage tool installation and lifecycle, this allows for the execution of unverified remote code if an untrusted repository is used.
[PROMPT_INJECTION]: The skill describes how asdf automatically reads and processes .tool-versions files found in project directories. This creates an indirect prompt injection surface where untrusted data within a repository could potentially influence the agent's behavior or the underlying system environment.
Ingestion points: .tool-versions files in current or parent directories, as specified in references/tool-versions.md.
Boundary markers: No delimiters or safety warnings regarding the content of .tool-versions files are present.
Capability inventory: The skill uses subprocess calls for tool management and network operations via git clone and asdf plugin commands (referenced in references/command-cookbook.md).
Sanitization: There is no evidence of sanitization or validation for the content of the version files before they are processed by the tool.

asdf