The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from the internet.
Ingestion points: The understandLink function fetches content from any URL provided at runtime (e.g., in SKILL.md).
Boundary markers: The documentation does not specify the use of delimiters or 'ignore' instructions to prevent the agent from obeying commands found within the fetched web content.
Capability inventory: The skill utilizes the node binary and the web-fetch skill to perform network and local execution tasks.
Sanitization: There is no evidence of content sanitization or filtering applied to the fetched HTML or text before it is passed to the NLP components.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to retrieve content from external, non-whitelisted domains as part of its core functionality using the web-fetch dependency.
[COMMAND_EXECUTION]: The skill executes local code using the node runtime to run its analysis script (analyzer.js), as seen in the usage and CLI examples.

link-understanding