The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from arbitrary web URLs, creating a vulnerability surface where malicious instructions embedded in web content could influence agent behavior.
Ingestion points: External web pages accessed via the <url> parameter in the defuddle parse command.
Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from following instructions found within the parsed web content.
Capability inventory: The skill documentation suggests an integration with a workflow (obsidian-source-ingestion) that writes data to local storage (Sources/Web/).
Sanitization: There is no mention of sanitization, filtering, or validation of the content extracted from the web pages before it is processed or stored.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a command-line tool, which is a standard functional requirement for its stated purpose.
Evidence: The usage of defuddle parse <url> --md in the provided documentation.

defuddle