The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to apply execution permissions using chmod +x to generated script files (start.command and start.sh) for cross-platform compatibility. It also configures the project's package.json to use the --host 0.0.0.0 flag, which causes the development server to listen on all available network interfaces.
[EXTERNAL_DOWNLOADS]: The generated startup scripts include npm install commands, which fetch external dependencies from the public npm registry. This is a standard procedure for the Vite-based development environment defined by the skill.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its requirement-gathering workflow.
Ingestion points: Untrusted user input is collected through a guided Q&A process in SKILL.md and references/portable-agent-prompt.md to define the project's subject, subject matter, and features.
Boundary markers: The skill does not define boundary markers or 'ignore embedded instruction' warnings when incorporating user responses into the generated design documents or source code.
Capability inventory: The skill has the ability to write multiple files to the local workspace and execute shell commands (e.g., npm, chmod).
Sanitization: There is no evidence of input validation or escaping logic to prevent a user from providing malicious requirements that could result in the generation of harmful code.

learning-assistant-builder