Use this skill when you need to retrieve, navigate, and synthesize evidence from the prebuilt LinearRAG textbook hierarchy across communities, docs, chapters, core sections, tiny sections, chunks, and local graph neighbors. Use it for topic routing, scoped drill-down, graph-based trace from strong anchors, and evidence-grounded textbook synthesis.
Use this skill when the task requires query-time retrieval over the prebuilt textbook hierarchy rather than offline graph construction.
This skill is for query-time retrieval, navigation, and evidence grounding over the prebuilt textbook hierarchy.
It is not for offline indexing, profiling, community construction, embedding rebuilds, or other preprocessing work.
If the user's prompt already contains the full local object needed to answer the question, such as a code snippet, formula, proof fragment, or exercise statement, use that prompt-local material first and use textbook retrieval only to corroborate or sharpen the answer.
Hard boundary:
Hierarchy:
community -> doc -> chapter -> core_section -> tiny_section -> chunkPractical roles:
chapter is the main routing layercore_section is the default middle anchorchunk is the final grounding layerUse five distinct tool families:
scripts/search_nodes.py
scripts/expand_structural.py
scripts/get_outline.py
scripts/expand_neighbors.py
scripts/fetch_content.py
Do not use other local scripts or other retrieval paths as substitutes for these tool families while operating under this skill.
Use the references in this order:
search_nodes.py and fetch_content.py I/O guaranteesReference split:
workflow.md
command-recipes.md
tool-contracts.md
search_nodes.pychapter for routing unless the query is already very specificcommunity as the default first hoplevelexpand_structural.py only for deterministic upward movesexpand_neighbors.py only after a useful anchor already existsreferences/workflow.md