The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables an agent to interact with databases based on user instructions, creating a surface where malicious user prompts could attempt to influence database operations.
Ingestion points: User instructions passed to the agent to perform database queries, migrations, or CLI operations as described in SKILL.md and references/cli-commands.md.
Boundary markers: The 'Netlify house rules' in SKILL.md provide explicit constraints, such as requiring user confirmation for destructive operations and mandating the use of migration files for production schema changes.
Capability inventory: The skill allows querying via @netlify/database and drizzle-orm, creating and applying migrations, and executing SQL commands through the netlify database connect CLI tool.
Sanitization: The documentation strongly advocates for parameterized queries via the sql template tag and provides explicit warnings in SKILL.md against using sql.raw for user-supplied input to prevent SQL injection.
[SAFE]: The skill documents the use of official, well-known libraries and tools from Netlify and established database community projects.
[SAFE]: Security best practices for secret management are correctly emphasized, instructing users and agents to use environment variables for connection strings instead of hardcoding credentials.
[SAFE]: The documentation includes detailed operational warnings regarding data privacy (PII in preview branches) and guides users to use appropriate access controls.

netlify-database