The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is designed to fetch article content from user-provided URLs. While this is necessary for its primary function, it involves making network requests to arbitrary external domains.
[COMMAND_EXECUTION]: The skill instructions include a 'Local render verification' step that executes shell commands. Specifically, it uses npx -y @mermaid-js/mermaid-cli to download and run the Mermaid CLI tool from the npm registry to verify diagram syntax.
[COMMAND_EXECUTION]: The skill mentions capturing screenshots using headless browsers like Playwright or Puppeteer if available, which involves executing browser automation tools.
[COMMAND_EXECUTION]: It performs file system operations such as writing temporary .mmd and .md files to the working directory and deleting them after use.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes the full text of untrusted articles provided via text or URL.
Ingestion points: Article text or content fetched from a URL.
Boundary markers: None explicitly defined to separate the untrusted article content from the agent's instructions.
Capability inventory: The skill has capabilities for shell command execution (npx), file system access (writing/deleting temp files), and network requests (fetching URLs).
Sanitization: The skill includes specific security rules for SVG generation, explicitly forbidding <script> tags, event handlers (onclick), and <foreignObject> to prevent XSS. However, it lacks robust sanitization for the inputs used to generate Mermaid or SVG content beyond standard escaping.

article-illustrate