The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves data from a remote curated index through the search_agents and get_site_details tools. While the index is curated, processing external site descriptions presents a standard surface for indirect injection.
Ingestion points: Search results and site metadata ingested from nothumansearch.ai in SKILL.md.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the interpolated data.
Capability inventory: The skill is restricted to remote JSON-RPC calls and does not perform local file writes or command execution.
Sanitization: No explicit sanitization or filtering of site content is described in the prompt logic.
[DATA_EXPOSURE]: The register_monitor tool requests a user email address. The skill instructions specifically mandate that this tool should only be used when a user explicitly provides an email and gives consent, which aligns with privacy best practices for handling PII.

not-human-search-mcp