The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input and interpolates it into shell commands, creating a surface for potential command injection or behavior override if inputs contain shell metacharacters.
Ingestion points: The <theme> argument in the generation command, the <instruction> argument in the edit command, and external file content via the -c, --context option in SKILL.md.
Boundary markers: None; user input is placed directly into command strings without delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill executes shell commands using npm, npx, or a local binary, providing a path to system-level execution.
Sanitization: Absent; there are no instructions to sanitize or escape user-provided strings before they are executed in the shell.
[COMMAND_EXECUTION]: The skill's primary operation involves executing shell commands. It specifically provides instructions to run code from a local path (/Volumes/Data/dev/imgen) using npx tsx or npm run, which gives the agent permission to execute logic within that local directory.
[SAFE]: The skill follows security best practices by recommending the use of environment variables and .env files for managing Azure OpenAI API keys, rather than hardcoding credentials.
[SAFE]: Usage of the @nogataka/imgen package and references to the developer's local environment are consistent with the skill's stated purpose and author identity.

imgen