The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill performs environment bootstrapping by downloading Node.js binaries from official sources (nodejs.org) and installing the vendor's package (@iqinghu/qhkit) via NPM. It includes a checksum verification step for the downloaded binary to ensure integrity.
[COMMAND_EXECUTION]: The skill constructs and executes shell commands for the qhkit tool, incorporating user-provided parameters such as prompts and model labels. It also instructions the agent to execute dynamic scripts for image processing using Python and Node.js.
[PERSISTENCE]: The instructions suggest modifying the user's shell profile (~/.bashrc) to persist the modified PATH environment variable across sessions, which is a common persistence mechanism used here for environment configuration.
[CREDENTIALS_UNSAFE]: The skill guides the user to provide an API token (QHKIT_TOKEN) and instructs the agent to store it using the CLI tool's configuration command or by setting environment variables.
[DYNAMIC_EXECUTION]: The skill uses python -c to execute inline Python code and npx to run Node.js packages (sharp-cli) dynamically for image compression tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (CLI tool outputs and user-supplied prompts) and interpolates them into commands, which presents a surface for potential instruction injection, although it implements a manual confirmation step before executing credit-consuming tasks.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from external sources including nodejs.org for the runtime and registry.npmmirror.com (a well-known Alibaba mirror) for packages.

linkpix-xiaohongshu-viral-video