The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides installation instructions that fetch software and dependencies from standard registries (npm, PyPI, Cargo) and the vendor's own repositories (GitHub, R-universe). These resources are managed by the author, 'xberg-io'.
[COMMAND_EXECUTION]: The html-to-markdown CLI tool supports a --url flag, which allows the agent to make network requests to fetch and process remote HTML content.
[PROMPT_INJECTION]: The skill serves as a gateway for processing untrusted external data, which introduces a surface for indirect prompt injection attacks.
Ingestion points: The skill accepts input from local files, stdin, and remote URLs (via the --url flag or SDK convert functions), as documented in SKILL.md and references/cli-reference.md.
Boundary markers: The documentation lacks specific instructions for agents to use delimiters or "ignore embedded instructions" warnings when interpreting the converted Markdown output.
Capability inventory: The library focuses on conversion and metadata extraction; it does not contain built-in functions for executing the resulting Markdown, which limits the risk of immediate exploitation.
Sanitization: The tool provides a --preprocess flag with adjustable presets (minimal, standard, aggressive) that can be configured to strip dangerous HTML tags like <script>, <style>, and <form> before conversion takes place.

html-to-markdown