html-to-markdown-using-the-mcp-server

Pass

Audited by Gen Agent Trust Hub on Oct 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The MCP launcher is designed to automatically acquire the html-to-markdown binary if it is not present on the system PATH or in the local cache. It attempts to resolve this via several channels, including downloading prebuilt binaries from the project's official releases on GitHub. These operations target well-known services and the vendor's own infrastructure.
  • [COMMAND_EXECUTION]: The skill operates by executing the html-to-markdown binary with the mcp subcommand to provide the MCP server interface for the agent. This execution is central to the skill's primary function.
  • [DYNAMIC_EXECUTION]: The bundled launcher uses dynamic resolution logic to locate or install the required binary at runtime. It checks local paths, utilizes package managers like Homebrew, npx, or uvx, and can fall back to remote downloads based on the HTML_TO_MARKDOWN_LAUNCHER environment variable.
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process HTML strings, which are often sourced from untrusted external websites.
  • Ingestion points: The html argument provided to both the convert_html and extract_metadata tools in SKILL.md.
  • Boundary markers: The documentation does not specify the use of delimiters or specific instructions for the agent to ignore content within the converted Markdown output.
  • Capability inventory: The skill executes a local conversion binary. While the MCP tools themselves are restricted from direct network or file system access, the agent may be influenced by instructions embedded in the converted text.
  • Sanitization: The tool provides robust sanitization capabilities, including strip_tags, exclude-selectors, and PreprocessingOptions which can remove scripts, styles, forms, and navigation elements from the input HTML before conversion.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 28, 2026, 03:25 PM