The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and execute the defuddle package from the official NPM registry. This is the core mechanism of the skill and uses a well-known, trusted package registry.
[COMMAND_EXECUTION]: The skill executes shell commands via the defuddle.sh script. The script includes safety measures such as set -euo pipefail for error handling and performs basic URL protocol validation (checking for http:// or https://) before execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external web content, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Untrusted content is fetched from remote URLs provided at runtime.
Boundary markers: The script prints output directly to stdout without explicit delimiters, though the agent's calling environment typically handles the separation of tool output.
Capability inventory: The skill itself does not have high-privilege capabilities (like file system writes or environment variable access) that could be easily exploited by malicious instructions in the Markdown output.
Sanitization: The conversion process from HTML to Markdown naturally strips out executable scripts and many other malicious web vectors.

defuddle