The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where untrusted data provided by upstream agents could influence the generation of executable database scripts.
Ingestion points: The skill ingests data requirements and architectural context through handoff signals such as BUILDER_TO_SCHEMA, ATLAS_TO_SCHEMA, and GATEWAY_TO_SCHEMA (documented in SKILL.md).
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands specified in the instructions for processing these ingested data requirements.
Capability inventory: The skill is designed to produce DDL (Data Definition Language) and migration scripts, which are inherently executable within database environments (documented in SKILL.md and reference files).
Sanitization: The provided instructions and reference files do not detail specific sanitization or validation routines for incoming requirements before they are incorporated into schema designs.
[EXTERNAL_DOWNLOADS]: The skill references and provide guidance on using several well-known and standard database tools and frameworks.
Evidence: References to industry-standard tools including Prisma, TypeORM, Drizzle, Atlas v1, pgroll, Liquibase 5, Alembic, gh-ost, and pt-online-schema-change are found in reference/migration-patterns.md and other documentation files. These are documented as standard practices for database professionals.

schema