The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation and Python script utilize curl and requests to interact with remote target URLs for identifying caching behavior and headers.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect injection as it fetches and processes HTTP responses from user-provided target servers.
Ingestion points: Data is ingested via requests.get() in scripts/agent.py when testing authenticated endpoints and static extensions.
Boundary markers: The script does not employ boundary markers or specific 'ignore' instructions for the content retrieved from target servers.
Capability inventory: The script includes the ability to search for PII indicators and generate a JSON report of findings, which is then presented to the agent context.
Sanitization: There is no explicit sanitization or filtering of the response body content before it is processed by the internal _check_pii method or included in the output report.

performing-web-cache-deception-attack