The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes shell commands via the kaipai CLI to perform image processing tasks, verify tool versions, and manage user authentication sessions.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the meitu-kaipai-cli package from the official npm registry. It also downloads user-specified remote image assets using the kaipai download command for local processing.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from remote URLs and tool outputs, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Remote image URLs provided by users and command-line outputs (stdout/stderr) returned by the vendor CLI and server.
Boundary markers: The skill includes an explicit instruction in references/runtime.md for the agent to ignore any instructions contained within server-returned text.
Capability inventory: Includes shell command execution through the vendor CLI, network access for asset synchronization, and local file system write operations.
Sanitization: The skill relies on the meitu-kaipai-cli tool's internal validation mechanisms for media files and utilizes standard shell argument handling for file paths.

kaipai-image-remove-watermark