Agent Skills
› kbanc85/claudia
› databases
databases
GitHub用于检查 Claudia 统一生产记忆数据库及其备份状态，包括身份、健康度、计数、模式等。仅在用户明确批准时进行数据库覆盖切换操作。
Trigger Scenarios
需要检查数据库身份或健康状态
需要查看数据库计数、模式或 WAL 信息
需要执行显式批准的数据库覆盖切换
Install
npx skills add kbanc85/claudia --skill databases -g -y
SKILL.md
Frontmatter
{
"name": "databases",
"description": "Inspect Claudia's unified production memory database and backups. Use for database identity, health, counts, schema, WAL, backup inventory, or an explicitly approved database override."
}
Claudia Databases for Codex
Inspect first; mutate only with explicit approval.
Default view
Prefer Claudia Memory tools when available. Otherwise inspect ~/.claudia/memory/claudia.db read-only and report:
- canonical database path and file size;
- counts for memories, entities, relationships, episodes, commitments, and sources;
- WAL/SHM presence;
- most recent verified backup;
- any legacy hash databases that still contain data.
Never infer health from file existence alone. Use memory-health for the complete system check.
Database overrides
Switch only when the user explicitly names or approves the target database.
- Resolve the exact target path and verify it is a readable SQLite database.
- Show the current and target paths and explain that a Codex restart is required.
- Back up
.mcp.json. - Add
CLAUDIA_DB_OVERRIDEtomcpServers.claudia-memory.envin the workspace.mcp.json. - Run
npx get-claudia codex .so the generated plugin MCP declaration stays aligned. - Start a new Codex session and verify the active database through a memory health call.
Never edit the installed plugin cache and never write directly to Claudia's SQLite database.
Version History
- 80253c0 Current 2026-08-20 06:21
