| name | schema |
| description | Use this skill when creating, inspecting, or migrating Retriever's SQLite schema. It is the source of truth for table layouts, path rules, custom field registry behavior, manual field locks, and schema-version behavior.
|
| metadata | {"version":"1.1.17"} |
Operates under retriever:routing. If the user's intent actually fits a different tier — another retriever:* skill, a Tier 2 slash, a Tier 3 tools.py subcommand, or (last resort) direct DB access — stop and re-route against the ladder before continuing.
Retriever Schema
Use this skill whenever a task touches:
- SQLite table definitions
- schema migrations
- relative path rules
- preview records
- chunk storage
- custom field registry entries
- manual-value overwrite protection rules
- structured filter validation rules
Required references
- Treat ../tool-template/src/10_core.py, ../tool-template/src/40_schema_runtime.py, and the generated ../tool-template/tools.py as the current source of truth.
- Use
python3 skills/tool-template/tools.py schema-version to confirm the current schema_version and tool_version.
- Read schema.md for background terminology only; if it conflicts with the source fragments or
schema-version, the code wins.
Rules
- Do not invent new table names when an existing table fits the need.
- Keep document paths relative to the workspace root.
- Keep preview paths relative to
.retriever/.
- Preserve manual edits by locking corrected document fields until explicitly overwritten.
- Prefer structured filter contracts over raw SQL filter strings.
- Make migrations idempotent.
- Prefer additive schema changes over destructive rewrites.