- name
- knowledge-router
- description
- Build a lightweight routing layer across existing knowledge sources such as MEMORY.md, daily memory files, self-improving notes, skill references, and audit records. Use when you need to decide where a question should be answered from before reading too much, when you want a unified knowledge index, or when you want to detect which knowledge should be promoted into a more durable layer.
Knowledge Router
Route queries to the right knowledge layer before doing broad reading.
Core workflow
- Scan known knowledge sources and classify them by role.
- Infer the query intent: rule, fact, method, evidence, or improvement.
- Recommend primary and secondary sources.
- Emit promotion hints when knowledge seems mature enough to move upward into a more durable layer.
Read references as needed
- Read
references/source-types.mdfor the knowledge source model. - Read
references/routing-rules.mdfor intent-to-source routing rules. - Read
references/promotion-rules.mdfor when knowledge should be promoted or extracted. - Read
references/report-format.mdfor the report structure. - Read
references/release-minimal.mdbefore packaging or publishing so the first public surface stays minimal.
Use scripts as needed
- Use
scripts/knowledge_router.py "<query>" [--scope ...] [--output report.txt]to build a routing report.
Operating rules
- Prefer routing over re-storing knowledge.
- Prefer a small number of clearly justified sources over broad search noise.
- Treat audit logs as evidence, not as the first answer source for general method questions.
- Keep the first version focused on source typing, query intent, and recommendation quality.