The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow for submitting articles uses the wxcli command with direct string interpolation of the user-supplied <title> and generated <html> content (e.g., wxcli draft add --title "<title>" --content "<html>"). Without rigorous sanitization or the use of safe execution APIs, this pattern is susceptible to command injection if the input contains shell metacharacters like backticks, dollar signs, or escaped quotes.
[EXTERNAL_DOWNLOADS]: The skill fetches and executes the markdown-to-html-cli package at runtime using npx. While the NPM registry is a well-known service, this creates a dependency on external code that is executed with the agent's privileges.
[PROMPT_INJECTION]: The skill processes untrusted user data (draft) to generate Markdown and HTML, creating an indirect prompt injection vulnerability surface.
Ingestion points: The draft and title inputs defined in SKILL.md are the primary entry points for untrusted data.
Boundary markers: The instructions do not define delimiters or specific "ignore embedded instructions" warnings when processing the draft.
Capability inventory: The skill has the ability to write files to /tmp and execute shell commands via npx and wxcli.
Sanitization: There are no explicit sanitization or escaping steps defined before the content is passed to the CLI tools or the shell environment.

wxarticle_creator