The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill describes processes for ingesting and executing user-supplied data, specifically through database seeding with a SQL file. This creates a potential surface for indirect prompt injection if the content of the file is influenced by untrusted sources.
Ingestion points: The skill references processing a user-provided seed.sql file in the CI workflow example within SKILL.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the provided command examples.
Capability inventory: The skill uses container execution commands (e.g., docker exec) to run database tools for provisioning and seeding, as detailed in the SKILL.md instructions.
Sanitization: There is no mention of sanitizing or validating the content of the SQL files before execution.
[Credential Management]: The skill correctly emphasizes the use of platform-native secrets (e.g., GitHub Actions secrets) for managing sensitive information like registry credentials and database passwords, which is a key security practice.
[Containerized Execution]: By executing database commands inside the service container, the skill avoids requiring the installation of database client tools on the CI runner, which helps maintain a minimal and more secure runner environment.

azuresql-db-ci