The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes results from database queries and schema inspections. Untrusted data stored within the database could potentially contain instructions aimed at influencing the agent's behavior during processing.
Ingestion points: Database query results from mssql sql and table schema details from mssql schema (SKILL.md).
Boundary markers: The skill includes explicit instructions for the agent to present full SQL queries to the user and explain their impact before executing dangerous operations.
Capability inventory: The skill can execute CLI commands, perform network operations to connect to databases, and write configuration files to the local file system.
Sanitization: No explicit programmatic sanitization of database results is defined; the skill relies on the agent's reasoning and mandatory user approval for modifications.
[COMMAND_EXECUTION]: The skill operates by constructing and executing shell commands via the mssql CLI. The agent is responsible for translating user requests into appropriate CLI arguments, including potentially complex SQL query strings.
[EXTERNAL_DOWNLOADS]: The skill documentation references the installation of the @billpeet/mssql-cli package from the official NPM registry. This package is a vendor resource corresponding to the skill author's namespace.

mssql