The Agent Skills Directory

[Indirect Prompt Injection] (SAFE): The skill describes a surface for processing untrusted data from database tables.
Ingestion points: Results returned from executeReadOnlySQL queries (defined in SKILL.md).
Boundary markers: Absent; there are no specific delimiters or instructions provided to the agent to differentiate between data and instructions within query results.
Capability inventory: The agent has access to executeWriteSQL (DML/DDL) and writeSecurityRule (permission management), which are high-impact capabilities.
Sanitization: Absent; the skill relies on the agent's internal reasoning and manual confirmation steps ('summarize what you are about to run') rather than technical sanitization.
[Command Execution] (SAFE): The skill provides the executeWriteSQL tool, which allows for the execution of arbitrary SQL commands including data modification and schema changes. This is identified as the primary intended purpose of the skill for database administration and management.
[Privilege Escalation] (SAFE): The skill provides writeSecurityRule, a capability to modify access control lists on database tables. This is an administrative function consistent with the skill's stated purpose of database management.

relational-database-mcp-cloudbase