The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of CLI tools (e.g., npx wavelet dev, npx wavelet push) and provides configuration for an MCP server running @risingwave/wavelet-mcp. These tools are for project initialization, database synchronization, and agent integration.
[PROMPT_INJECTION]: Documentation includes MCP tools such as run_sql and query_view which ingest data from external databases. This creates a surface for indirect prompt injection if the database contains untrusted data. • Ingestion points: Database query results via query_view and run_sql in SKILL.md. • Boundary markers: None identified. • Capability inventory: run_sql, emit_event, emit_batch, and CLI sync commands. • Sanitization: Supports JWT-based filterBy for enforcing row-level security.
[CREDENTIALS_UNSAFE]: The configuration examples use placeholder connection strings (e.g., postgres://user:pass@host:5432/db) and correctly recommend the use of environment variables for sensitive data like JWT_SECRET and POSTGRES_URL.

wavelet