fux-enrich

Generates doc2query enrichment questions for documents in a Fux search corpus.

1|Updated Jun 1, 2026
One-click install
npx skills add https://github.com/arpitarya/fux --skill fux-enrich-arpitarya
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fux-enrich
Source: https://github.com/arpitarya/fux/tree/main/.agents/skills/fux-enrich
Command: npx skills add https://github.com/arpitarya/fux --skill fux-enrich-arpitarya

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Documents in a Fux search index can only be found by words they literally contain, so a postmortem titled "checkout unavailable for 47 minutes" will never match a search for "what happened during the checkout outage". This Skill writes the questions a searcher would type, attaching new searchable vocabulary to each document so it can be found by words it does not contain. ## Core Features & Use Cases - Question generation (doc2query): Writes five to ten natural-language questions per document into .fux/enrich/<sha>.md files that are committed to git and indexed as searchable context. - Plan-driven workflow: Uses fux enrich --plan to get the exact worklist, sha, and chunk count, then re-verifies the sha immediately before writing to avoid stale enrichment. - Retrieval validation: Every question is checked with fux enrich --check, which runs each question as a real search and refuses the file if the question does not place its own document in the top 3. - Use Case: A user asks to "enrich the ranking SR". The Skill runs the plan, reads records/0111_ranking.md, writes questions like "How does fux decide which result comes first?", validates them, and the document becomes findable by searcher vocabulary. ## Quick Start Enrich the document records/0111_ranking.md in my Fux corpus so it can be found by questions it does not literally answer.

Frequently Asked Questions about fux-enrich

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I enrich a document in a Fux search index?▼

Run fux enrich --plan with the document path to get its sha and chunk count, read the document, then write five to ten searcher-vocabulary questions into .fux/enrich/<sha>.md with the required frontmatter. Validate the result with fux enrich --check before finishing.

What is doc2query enrichment for search ranking?▼

Doc2query is a technique that generates questions a document could answer and indexes them as extra searchable text, so the document matches queries using words it never contains. Fux applies the doc2query-- filter, refusing any generated question that does not retrieve its own document in the top 3 results.

Why does fux enrich --check refuse my enrichment file?▼

The check runs every question as a real search with the title and ctx fields zeroed, and refuses the file if a question fails to place its own document in the top 3. It also refuses files missing required frontmatter keys or containing patterns matched by rules in .fux/pii.toml.

What does a STALE status mean in fux enrich?▼

STALE means the document changed after it was last enriched, so the recorded sha no longer matches. Treat it like a missing enrichment: re-run the plan, write a fresh file under the new sha, and do not edit the old file.

When should I not run fux enrichment?▼

Only run it when a human explicitly asks to enrich, re-enrich, or fill in enrichment, because the generated files are committed to git and change search ranking. Never run it as a side effect of another task, and ask before proceeding when a plan covers multiple documents without a named scope.