The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation in reference/shared.md instructs the user or agent to install the @volcengine/mediakit-cli package globally via the NPM registry. This is a vendor-provided tool required for the skill's functionality.
[COMMAND_EXECUTION]: All image processing tools (e.g., image-ocr, erase-image, enhance-image) are implemented by executing the mediakit-cli binary through the shell. The SKILL.md frontmatter explicitly requests shell permissions.
[DATA_EXFILTRATION]: The skill is designed to send image data (via URLs) and metadata to Volcengine's cloud infrastructure for processing. It also manages authentication via the MEDIAKIT_API_KEY environment variable and stores configuration in ~/.mediakit/config.json.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection.
Ingestion points: Untrusted external data, such as image_url, callback_args, client_token, and standard_erase_text, is accepted as parameters for CLI commands.
Boundary markers: There are no instructions to use delimiters or ignore embedded instructions when constructing the shell commands.
Capability inventory: The skill has the capability to execute shell commands with user-supplied arguments (documented in reference/shared.md and individual tool files).
Sanitization: There are no explicit instructions for the agent to sanitize, escape, or validate user-provided strings before they are interpolated into the shell execution template.

byted-mediakit-image