The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads Node.js binaries and verification checksums directly from nodejs.org (a well-known service) to perform automated environment setup when the runtime is missing.
[PERSISTENCE]: Instructions direct the agent to append the installation path to ~/.bashrc, which ensures the tool remains available across different shell sessions.
[DYNAMIC_EXECUTION]: The skill contains a Python snippet using the PIL library and a Node.js sharp-cli command that are dynamically executed to compress images if they exceed the 10MB size limit.
[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted user input that is interpolated into shell commands, creating a potential surface for indirect injection.
Ingestion points: User-provided customCopy (up to 500 characters) and prompt fields in SKILL.md.
Boundary markers: The inputs are wrapped in JSON strings within shell commands, but no explicit escaping or boundary markers are defined to prevent command injection.
Capability inventory: Execution of qhkit CLI, python scripts, and npm/npx installation commands.
Sanitization: There is no evidence of sanitization or validation logic for the input strings before they are passed to the shell.
[COMMAND_EXECUTION]: The skill frequently invokes shell commands for package management (npm), environment configuration, file manipulation (tar, sha256sum), and primary tool functionality.

linkpix-shopee-image