The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs shell commands using string interpolation of user-provided data. The bash template in the Execute section uses the user-supplied <image_url> and <target_background_description> directly within a command line. This pattern is susceptible to shell injection attacks if the input contains shell metacharacters like semicolons, backticks, or command substitution sequences.
[CREDENTIALS_UNSAFE]: The skill is configured to read sensitive authentication data from local files. It requires access to ~/.meitu/credentials.json and uses environment variables MEITU_OPENAPI_ACCESS_KEY and MEITU_OPENAPI_SECRET_KEY to authenticate CLI operations.
[EXTERNAL_DOWNLOADS]: The skill documentation references the installation of external software from a remote package registry. It provides instructions to install the meitu-cli tool using the npm package manager.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection. Ingestion points: Data enters the system via the user-provided prompt and image_url parameters. Boundary markers: Double quotes are present in the bash template, but there are no instructions to sanitize or escape characters that could break the string boundary. Capability inventory: The skill has the authority to execute CLI commands and write to local output directories. Sanitization: There is no mention of input validation, filtering, or escaping for the user-supplied strings before they are processed.

image-background-replace