The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to download and install Node.js from nodejs.org if it is missing from the environment. It also installs the vendor's CLI tool @iqinghu/qhkit from the official NPM registry or a trusted mirror (registry.npmmirror.com). These are well-known, trusted services.
[COMMAND_EXECUTION]: The skill relies on executing the qhkit CLI to perform image generation, configuration, and estimation tasks. It also uses python (with the pillow library) or npx (with sharp-cli) to locally compress images that exceed size limits. These operations are core to the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill represents a potential surface for indirect prompt injection as it ingests untrusted user data which is then processed by external tools.
Ingestion points: The prompt and uploadedImages parameters in SKILL.md are the primary entry points for external data.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to isolate or ignore potentially malicious instructions embedded within the user's prompt or image metadata.
Capability inventory: The skill possesses capabilities for shell command execution (qhkit, npm, python), file system writes (image compression), and network access (via the CLI tool).
Sanitization: No specific sanitization or validation logic is defined for the input prompts or image files before they are passed to the underlying tools.

linkpix-seedream-5-pro