Ridge DB for AI Agents
The ridgedb skill gives LLMs and coding agents the operational Ridge
capability map, supported SQL, connection choices, safety boundaries, Ridge
Admin workflows, and deterministic read-only inspection scripts.
Install
Download the current repository skill package:
Extract it into the agent's skill directory or keep the repository copy at:
.agents/skills/ridgedb/
The skill contains SKILL.md, UI metadata, focused references, and read-only
probe and documentation-drift scripts.
Trigger examples
Use $ridgedb to inspect the local database and list its schemas and tables.
Use $ridgedb to write a Ridge-compatible migration for this application.
Use $ridgedb to diagnose this pgwire SQLSTATE without assuming PostgreSQL.
Use $ridgedb to plan and verify a non-overwriting backup restore.
Use $ridgedb with $seed-lang to embed Ridge in this seed application.
Agent safety contract
The skill requires agents to:
- inspect version, endpoint, and catalog before generating SQL;
- prefer read-only discovery;
- use Ridge-native
SHOWandDESCRIBEcommands; - distinguish embedded Ridge,
ridged, pgwire, and Ridge Admin; - obtain explicit authorization for destructive DDL or operational mutation;
- never restore over an existing target;
- never edit internal database files;
- keep network services on loopback and use verified SSH remotely;
- treat lost write acknowledgments as uncertain and never auto-replay them;
- state unsupported behavior instead of inventing PostgreSQL compatibility.
Machine-readable discovery
The site publishes a compact llms.txt index. Agents working from the repository should prefer the canonical capability and release documents named by the skill, because they track the implementation revision directly.
For system-level work, begin with the runtime architecture and use its interactive Archify map to trace the authored request, SSH, durability, and security paths. The map is an orientation aid; current code, contracts, and focused tests remain authoritative.