The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a monitoring tool that fetches and parses JSON data from external vendor status pages. This creates a surface for ingesting untrusted content.
Ingestion points: The check_status_page function in SKILL.md fetches data from /api/v2/summary.json endpoints.
Boundary markers: The script uses standard JSON parsing (json.loads) and retrieves data into structured Python dictionaries.
Capability inventory: The script uses the requests library for network retrieval but does not invoke subprocesses or perform file writes.
Sanitization: The script limits the response body read to 256KB (262145 bytes) and documentation explicitly instructs administrators to allowlist trusted hosts and reject metadata or private destinations.

vendor-management