The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests metadata and SQL definitions from the Materialize system catalog to generate reports, creating a potential surface for indirect prompt injection if those catalog objects contain malicious content.
Ingestion points: Queries targeting mz_internal ontology views and mz_catalog objects in SKILL.md and references/queries.md.
Boundary markers: The instructions do not specify delimiters (e.g., XML tags or backticks) to wrap external database output to prevent the agent from interpreting it as instructions.
Capability inventory: The skill uses query_system_catalog and query tools for database access and provides shell-based state probes.
Sanitization: No explicit logic is provided to sanitize or validate database output before it is included in the final report template.
[COMMAND_EXECUTION]: The skill defines several shell-based "state-detection probes" in mcp-client-connect.md (e.g., using curl, psql, and claude mcp) that the agent is intended to run to verify environment configuration during setup.
[EXTERNAL_DOWNLOADS]: The skill references the official Materialize documentation and API endpoints on materialize.com, which are legitimate vendor resources.

mz-health-check