The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the meitu CLI with image_url and prompt parameters. If these inputs contain shell metacharacters, it could lead to unintended command execution as no explicit sanitization is documented for these tool calls.
[EXTERNAL_DOWNLOADS]: The skill includes a preflight step that triggers a global installation or update of the meitu-cli package using npm install -g if the local version is outdated.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external image URLs and user-provided subject descriptions, which are used to generate instructions for the vision processing API. Ingestion points: image_url and prompt parameters in SKILL.md. Boundary markers: None identified. Capability inventory: Execution of meitu CLI, file reads from ~/.meitu/, and file writes to ./output/. Sanitization: No explicit sanitization or escaping of the input parameters is mentioned in the workflow.
[CREDENTIALS_UNSAFE]: The skill is designed to read sensitive authentication data from ~/.meitu/credentials.json to authenticate its CLI operations with the Meitu OpenAPI.

image-cutout