The Agent Skills Directory

[PERSISTENCE]: The instructions in SKILL.md guide the user to modify their shell startup configuration by appending an environment variable export to ~/.zshenv (echo 'export GOOGLE_CLOUD_PROJECT=...' >> ~/.zshenv). While this is used for legitimate Google Cloud authentication, it represents a persistent modification to the user's shell environment.
[DYNAMIC_EXECUTION]: The scripts/upscale.ts script uses dynamic loading to resolve dependencies. It programmatically searches for utility scripts in various locations, including environment-defined paths and internal platform plugin directories (e.g., ~/.claude/plugins/), and imports them using computed paths.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data which could potentially be used as an injection vector.
Ingestion points: The skill accepts a user-provided image file path as a command-line argument and reads the binary data of that image in scripts/upscale.ts.
Boundary markers: No explicit delimiting or instructions to ignore embedded content were found.
Capability inventory: The skill has the capability to write files to the local system using the saveImage function.
Sanitization: No specific sanitization or integrity verification is performed on the input image or the provided file path before processing.

upscale-image