The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains instructions for the agent to install prerequisites by downloading a shell script from a remote URL (https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh) and piping it directly to bash. This source is not included in the trusted vendor list, representing an unverifiable remote code execution risk.
[COMMAND_EXECUTION]: The skill's troubleshooting documentation (SKILL.md) directs the agent to modify the internal structure of the global @adobe/aio-cli installation by performing a manual npm installation inside the global package directory to resolve module resolution issues. This represents high-privilege modification of shared system tools.
[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data during the project initialization process.
Ingestion points: User-provided inputs for template names, project paths, and environment identifiers are processed at runtime.
Boundary markers: No explicit delimiters or warnings to ignore embedded instructions are provided for user-supplied values.
Capability inventory: The skill has access to the bash shell, file system write operations, and the ability to modify project configuration files (app.config.yaml, ext.config.yaml).
Sanitization: While the init.sh script uses bash arrays to prevent basic command injection, there is no validation of the source or content of the template identifiers provided by the user.
[EXTERNAL_DOWNLOADS]: The skill triggers downloads of external scripts and packages. While Adobe-scoped packages originate from the vendor, the inclusion of the nvm installer from an external repository constitutes a non-trusted download within an execution context.

appbuilder-project-init