The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes subcommands of the meitu CLI tool (such as image-cutout, image-edit, and image-upscale) to perform its core image processing tasks.
[CREDENTIALS_UNSAFE]: Accesses sensitive API credentials from the environment variables MEITU_OPENAPI_ACCESS_KEY and MEITU_OPENAPI_SECRET_KEY, as well as the file ~/.meitu/credentials.json. This access is documented as a requirement for authenticating with the vendor's API.
[EXTERNAL_DOWNLOADS]: Directs users to install the meitu-cli package from the NPM registry. Since this resource belongs to the skill author ('meitu'), it is categorized as a legitimate vendor dependency.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted user data into its internal prompts.
Ingestion points: User-provided text descriptions and image analysis results enter the agent context in SKILL.md (Step 1).
Boundary markers: Delimiters or instructions to ignore embedded commands are absent in the references/prompts.md templates.
Capability inventory: The skill possesses command execution capabilities using the meitu tool as specified in SKILL.md (Steps 4-7).
Sanitization: No validation or escaping of user-supplied content is performed before interpolation into generation prompts.

meitu-product-view