The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill references several markdown files hosted on GitHub within the planetscale organization. While PlanetScale is a well-known database provider, the organization is not included in the predefined trusted sources list, making these external references unverifiable by the system's strict trust policy.
[COMMAND_EXECUTION] (LOW): Recommends the installation of pg_activity via pip install. In a strict 'assume-malicious' posture, installing third-party packages from non-whitelisted sources is flagged as a minor concern.
[COMMAND_EXECUTION] (SAFE): The skill provides numerous SQL queries and pscale CLI commands for database management. It demonstrates security best practices by including explicit bold warnings (e.g., 'Always confirm with a human before...') for all actions that could lead to data loss or service disruption.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill provides an attack surface for indirect prompt injection through the processing of database metadata and API responses.
Ingestion points: SQL query results from system views (pg_stat_activity, pg_stat_statements) and PlanetScale API response data.
Boundary markers: Not explicitly defined in the provided markdown templates.
Capability inventory: Subprocess execution (pscale CLI) and SQL execution capabilities.
Sanitization: No explicit sanitization of external content is described in the reference material.

postgres