The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of the xparse-cli binary via the host shell. To mitigate risks associated with command execution, the instructions explicitly forbid unsafe shell practices such as piping, heredocs, or inline JSON arguments. It requires creating temporary files with restricted 0600 permissions to pass task context, ensuring data is not exposed in process lists or logs.
[INDIRECT_PROMPT_INJECTION]: As a document parsing skill, it naturally processes untrusted data from user-provided files and instructions.
Ingestion points: Document files (PDF, Image, Office) and user intent instructions are passed to the CLI via the --files and --instruction flags in SKILL.md.
Boundary markers: The skill does not define specific prompt delimiters for the instruction field but relies on the underlying service's processing logic.
Capability inventory: The skill has the capability to execute shell commands, read/write local files, and perform network operations via the vendor's CLI tool.
Sanitization: The agent is instructed to pass user intent verbatim to the service. While this creates a potential surface for injection, the skill includes robust error-handling gates to detect and respond to service-level failures or unexpected inputs.
[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for installing the xparse-cli tool using the Node Package Manager (npm). The references include the standard NPM registry and the npmmirror.com registry (a well-known mirror for the China region). These are official distribution channels for the vendor's utility.

xparse-parse