The Agent Skills Directory

[SAFE]: The skill implements best practices for secret management. It contains explicit instructions for the agent to never request or handle sensitive information like SERVICE_ROLE_KEY or DATABASE_URI within the chat session. Instead, it guides the user to manage these credentials locally in gitignored environment files (e.g., .env.db, .env.secrets).
[EXTERNAL_DOWNLOADS]: The skill fetches a canonical model registry in JSON format from a remote repository (https://gist.githubusercontent.com/mhassanch/5acd83c04618c83e29d118ac722bb805/raw/geobase_models_registry.json). This file is used for configuration and validation of model parameters.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external URL to configure model workflows. * Ingestion points: Fetches JSON configuration from a GitHub Gist. * Boundary markers: None explicitly mentioned for the parsing step. * Capability inventory: The skill instructions facilitate command execution through the geobase-cli and standard data tools like psql and ogr2ogr once the environment is configured. * Sanitization: The skill requires the agent to validate the registry data, mapping user input to specific IDs and enforcing constraints defined in the JSON schema.

geobase