The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install the package @tinify-dev/client from the NPM registry. This is not the official library for the Tinify service, which typically uses the 'tinify' package name.
[REMOTE_CODE_EXECUTION]: Employs 'npx -y' to download and execute code directly from the NPM registry. Using the '-y' flag bypasses confirmation prompts, allowing for immediate execution of potentially malicious or unverified code at runtime.
[COMMAND_EXECUTION]: Executes shell commands including 'npx' and 'curl'. The curl example includes a subshell execution '$(uuidgen)', which triggers a system utility based on the skill's instructions.
[DATA_EXFILTRATION]: Configures the agent to send image data to 'https://api.tinify.dev'. This domain is not the official API endpoint for the Tinify service (api.tinify.com), potentially leading to the exposure of sensitive image data to an untrusted third party.
[PROMPT_INJECTION]: The skill processes external image files, creating an attack surface for indirect prompt injection. Ingestion points: User-provided image files (AVIF, WebP, JPEG, PNG, GIF). Boundary markers: None present to delimit untrusted data from instructions. Capability inventory: Includes network exfiltration (curl) and external code execution (npx). Sanitization: No explicit sanitization of image metadata or content is performed before processing.

tinify-images