The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires a global installation of the defuddle package via npm install -g defuddle. This introduces an unverified external dependency from a public registry that is not associated with a known trusted vendor.
[COMMAND_EXECUTION]: The skill executes the defuddle CLI using a user-supplied URL as a parameter (defuddle parse <url>). Directly passing external input to host-level commands presents a potential risk of command injection if the input is not properly sanitized.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process arbitrary web content, which can be used as a vector for malicious instructions targeting the agent.
Ingestion points: The usage of defuddle parse <url> in SKILL.md allows for the ingestion of data from any external URL.
Boundary markers: The skill lacks any instructions or delimiters to warn the agent that the fetched content should be treated as untrusted data or to ignore embedded commands.
Capability inventory: The skill possesses the ability to install software via npm and execute system-level commands on the host.
Sanitization: No sanitization, filtering, or validation steps are defined for the extracted markdown content before it is processed by the agent.

defuddle