The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file SKILL.md contains a hardcoded password string ('secret') within a SQL command example for creating a database replication user (CREATE ROLE replicator WITH REPLICATION LOGIN PASSWORD 'secret';). Providing literal credentials in setup documentation or templates can lead to insecure production deployments if users do not replace them with secure, unique passwords.
[COMMAND_EXECUTION]: The skill facilitates the execution of high-privilege shell commands (pg_basebackup) and administrative SQL operations (e.g., slot creation, role creation). While these are controlled by enums for specific operations, the skill directly interacts with the database system and underlying file system based on agent-interpreted tasks.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface for processing natural language tasks related to database replication, creating a surface for indirect injection if untrusted data is processed by the agent using these tools.
Ingestion points: Parameter values for operation and replication_mode defined in SKILL.md.
Boundary markers: No specific delimiters or instructions to ignore instructions embedded within data are provided in the skill context.
Capability inventory: The skill allows the agent to create database roles, manage replication slots, and perform system-level backups via pg_basebackup.
Sanitization: While operation and replication_mode utilize enums to restrict inputs, there is no evidence of broader context sanitization.

postgresql-replication