The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/segment.ts utilizes dynamic module loading with computed paths. It resolves the PLUGIN_ROOT by checking the GEMSKILLS_ROOT environment variable and reading the user's plugin configuration file at ~/.claude/plugins/installed_plugins.json. This mechanism allows for the execution of code located at paths determined at runtime.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. 1. Ingestion points: The skill accepts an image path and a user-provided prompt string via the --prompt flag in scripts/segment.ts. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded content are used when passing data to the vision model. 3. Capability inventory: The script has the capability to create directories and write files to the local filesystem using mkdir and writeFile in scripts/segment.ts. 4. Sanitization: The script performs regex-based sanitization on output filenames to remove non-alphanumeric characters, but does not sanitize the prompt or image data itself.

segment-image