The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use the Bash tool with curl to access web.archive.org, noting that the platform's native WebFetch tool is restricted for this domain. This is a functional workaround for data retrieval.
[EXTERNAL_DOWNLOADS]: The skill fetches data from web.archive.org. This is a well-known and legitimate service for historical web data. The operations do not involve sensitive local files or credential exfiltration.
[PROMPT_INJECTION]: The skill creates an indirect prompt injection surface because it fetches and processes untrusted HTML content from external archived websites.
Ingestion points: Data is ingested via curl commands in Bash and the fetch() function in the provided Python script.
Boundary markers: The instructions do not define explicit boundary markers to separate retrieved web content from the agent's instructions.
Capability inventory: The agent has access to the Bash tool (curl) and the ability to write to the local filesystem (e.g., scripts/.crawl_progress.json for tracking progress).
Sanitization: The skill provides basic parsing examples using regular expressions (e.g., title extraction) but does not implement comprehensive sanitization of the fetched HTML content before it might be interpreted by the agent.

wayback-machine-cdx-crawling