The Agent Skills Directory

[SAFE]: The skill uses a defined parameter schema in SKILL.md with a regex pattern (^[a-z][a-z0-9_]*$) for role names, which helps prevent SQL injection at the parameter level.
[DYNAMIC_EXECUTION]: The validation script scripts/validate.py correctly uses yaml.safe_load() when parsing the configuration file, preventing potential remote code execution (RCE) vulnerabilities associated with unsafe YAML deserialization.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided task descriptions to perform database administration tasks. This represents a standard input surface for the intended functionality.
Ingestion points: User task descriptions provided via the CLI (e.g., `claude "postgresql-admin
[task]"`).
Boundary markers: None explicitly defined in the usage examples.
Capability inventory: Database role management and security auditing (e.g., CREATE ROLE, GRANT, REVOKE).
Sanitization: Input validation is applied to the role_name parameter via regex in the skill definition.

postgresql-admin