The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted HTML data from external sources, creating an attack surface where malicious instructions embedded in the HTML could be delivered to the agent after conversion to Markdown.
Ingestion points: External HTML is ingested via the --url CLI flag and the convert() library functions across all supported language bindings (Python, Rust, Node.js, Go, Ruby, PHP, Java, C#, Elixir, R).
Boundary markers: The documentation does not specify the use of delimiters or instructions to the agent to disregard natural language commands found within the converted Markdown output.
Capability inventory: The skill can perform network requests (--url) and write files to the local system (--output).
Sanitization: The tool performs structural sanitization by converting HTML to Markdown (e.g., removing scripts), but it does not sanitize the resulting text content for potential prompt injection vectors.
[EXTERNAL_DOWNLOADS]: The skill documentation describes installation procedures using standard package managers and official vendor repositories.
Evidence: Mentions installation via Homebrew (xberg-io/tap), pnpm (@xberg-io/html-to-markdown-cli), uvx (html-to-markdown), and Cargo (html-to-markdown-cli).
Evidence: The R package is sourced from the vendor's repository at https://xberg-io.r-universe.dev.
[COMMAND_EXECUTION]: The CLI tool includes a feature to perform network operations to retrieve data from external URLs.
Evidence: The --url flag allows the tool to fetch HTML content from a remote server for processing.

html-to-markdown-extracting-metadata