fux-inspect

What problem does it solve? Teams maintaining a Fux search index cannot tell whether the index is healthy — which documents no query can reach, which words appear on everything, or whether the corpus is full of near-duplicates. This Skill reads the shape of a Fux index with fux inspect and turns its report into actionable findings. ## Core Features & Use Cases - Index health checks: Reads the checks array first — unreachable share, boilerplate share, near-duplicate share, and the descriptive findable share — with provisional floors reported as signals, never failures. - Corpus diagnostics: Surfaces boilerplate terms, template families, near-duplicate pairs, undecodable or unreadable documents, graph orphans and hubs, and vocabulary statistics like Zipf slope and distinct-term percentiles. - Lever mapping without applying it: Every finding names its remedy (stopwords, .fuxignore, fux enrich, decoders) but the Skill is strictly read-only and offline — it proposes commands and never runs them. - Use Case: When a user asks "why can't fux find this document" or "is our index any good", run fux inspect --json, read findability.unfindable and the counts beside capped lists, and report the findings with their levers. ## Quick Start Ask the assistant to run fux inspect on this repository and summarize whether the index is healthy, listing any unfindable or duplicate documents.