The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download and execute code from a third-party GitHub repository (ericzakariasson/pg-mcp-server.git) and via the npx pg-mcp-server command. These sources are not verified or managed by a trusted organization.
[COMMAND_EXECUTION]: Through the query tool, the skill provides a mechanism for the agent to execute arbitrary SQL queries. This creates a significant attack surface where an agent could be manipulated into performing unintended database operations.
[DYNAMIC_EXECUTION]: The skill includes an optional configuration DANGEROUSLY_ALLOW_WRITE_OPS. If enabled, this grants the agent the ability to modify or delete data within the database (INSERT, UPDATE, DELETE), increasing the risk of data loss or corruption via agent manipulation.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves data from an external database and injects it into the agent's context, creating a vulnerability to indirect prompt injection.
Ingestion points: Data retrieved from database rows and schema information via the query tool and postgres:// resources are fed directly into the LLM context.
Boundary markers: The instructions do not define boundary markers or provide safety guidelines to help the agent distinguish between data and instructions within the query results.
Capability inventory: The skill possesses full read capabilities and optional write capabilities for the connected database.
Sanitization: There is no evidence of sanitization or filtering of the database content before it is processed by the LLM.

postgres-mcp-server