full-suite
full-suite is the top of the stack: it runs
concept-family-explorer to map a subject’s family,
researches every worthwhile gap to saturation, and then — where the family-explorer alone
would stop at a mapped tree — compiles the result into the full llms-family set (llms.txt,
llms-full, llms-small, llms-facts, llms-vocabulary) at both the per-concept and rollup
level, with keyword (FTS5) and semantic (embedding) indexes, and registers everything in
every reachable concept tree.
A cheap, low-effort frontrunner scout pre-checks candidate domains for an existing
llms.txt/llms-full.txt before the expensive passes run, seeding both /dr and
concept-family-explorer’s warm-start cache so the full run doesn’t re-derive what a target
site already publishes.
Use it for: “do a full suite on X”, “fully exhaust X”, “give me everything on X” — when the deliverable is the complete artifact set, not just a map or a report.
Not for: one named concept with no family-mapping needed (/dr) · a concept map with no
llms packs wanted (concept-family-explorer) · fixing one existing llms file
(llms-deep-optimizer) · abstracting one concept out of a
corpus you already hold (llms-concept-abstractor) · a
cited report with no artifacts (deep-research) · rebalancing an existing tree with no new
research (skill-tree-architect).