functional-area-resolver

To compress oversized LLM routing files, convert per-skill routing rows in RESOLVER.md or AGENTS.md into functional-area dispatchers. This replaces many granular entries with one dispatcher per area, preserving sub-skill reachability while shrinking the context budget.

A functional-area dispatcher is a single routing entry that replaces many granular skill-per-row entries in a large routing file. It uses a "(dispatcher for: ...)" clause so an LLM can still reach the correct sub-skill under limited context budgets.

You need to compress an AGENTS.md or RESOLVER.md routing table when the file exceeds approximately 12KB and begins wasting prompt tokens on pipe-table routing. This context-budget blowup occurs when hundreds of skill-per-row entries grow too large.

To verify routing accuracy after compressing an LLM routing file, run structural routing evaluation followed by an LLM harness A/B test on the edited file. This ensures the functional-area dispatchers preserve sub-skill reachability before committing changes.

Preconditions for compressing a large routing file require the file to be sufficiently large (over ~12KB) and the working tree to be clean unless explicitly forced. Gatekeeping these preconditions ensures safe functional-area dispatcher conversion.

Yes, functional-area dispatcher routing works with always-on LLM workflow entries by keeping them separate from the compressed dispatcher entries. This maintains their immediate availability while still shrinking the overall routing table footprint.