The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted HTML data from files or remote URLs (using the --url flag). This creates a surface for indirect prompt injection if the source content contains malicious instructions intended to influence an agent reading the converted Markdown output.
Ingestion points: Processes HTML from local files (stdin/FILE) or remote URLs specified via the --url parameter in references/cli-reference.md.
Boundary markers: The skill converts input to standard Markdown/Djot/Plain text but does not include explicit instructions for agents to ignore embedded commands in the output.
Capability inventory: The conversion logic is implemented in Rust, Python, and Node.js. The tool can fetch remote data, write to local files, and execute shell completions.
Sanitization: The skill provides preprocessing flags (--preprocess, --strip-tags, --exclude-selectors) to remove potentially dangerous or noisy HTML elements like scripts, forms, and navigation before conversion.
[EXTERNAL_DOWNLOADS]: The documentation provides multiple methods for installing the tool from external sources, including Homebrew, npm (pnpm), PyPI (uvx), and Crates.io (cargo).
Evidence: Instructions in references/cli-reference.md suggest brew install xberg-io/tap/html-to-markdown, pnpm dlx @xberg-io/html-to-markdown-cli, and cargo install html-to-markdown-cli.
[COMMAND_EXECUTION]: The skill involves the execution of shell commands to perform conversions and manage the CLI tool.
Evidence: SKILL.md and references/cli-reference.md provide numerous examples of bash commands for file processing, piping stdin, and tool configuration.
[REMOTE_CODE_EXECUTION]: The skill uses tools like pnpm dlx and uvx which fetch and execute code from remote registries at runtime.
Evidence: pnpm dlx @xberg-io/html-to-markdown-cli --help and uvx --from html-to-markdown html-to-markdown --help are documented in references/cli-reference.md.

html-to-markdown-converting-html