The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text for database search and query operations, which constitutes a potential attack surface.
Ingestion points: User-provided query strings and parameters (queryText, $1, $2) are used within SQL statements across SKILL.md and the search-specific reference files.
Boundary markers: The skill explicitly recommends managing migrations as code and using ORM-managed environments rather than ad-hoc SQL execution.
Capability inventory: The skill uses tools like run_sql, execute, and npx drizzle-kit to perform database modifications and queries.
Sanitization: The instructions mandate the use of parameterized queries and Drizzle's sql template tags to ensure that external inputs are handled safely.
[EXTERNAL_DOWNLOADS]: The skill references and installs additional configurations and standard development dependencies.
Evidence: Recommends fetching skill definitions from neon.com and installing official skills using the neon skills CLI.
Evidence: Instructs on installing well-known Node.js packages such as pg, dotenv, and @neon/ai-sdk-provider from the official registry.
Note: All identified downloads and references originate from established organizations or well-known services directly related to the skill's primary function.

neon-postgres