The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and return raw content from external URLs, which provides a vector for indirect prompt injection where malicious site content could influence agent behavior.
Ingestion points: Content ingested from external websites via the url parameter in POST /v1/scrape.
Boundary markers: The skill does not implement delimiters or safety instructions to distinguish data from instructions when processing the scraped content.
Capability inventory: The skill has access to sensitive tools including Bash (curl/node), Read, Write, Edit, and Grep.
Sanitization: No sanitization or filtering of the scraped data is performed before returning it to the agent.
[DATA_EXFILTRATION]: The skill instructions involve reading a local credential file at ~/.xcrawl/config.json and sending the extracted XCRAWL_API_KEY to the vendor's endpoint at https://run.xcrawl.com. This behavior is consistent with the intended functionality provided by the vendor.

xcrawl