resolver-querylisted

# /resolver-query Look up which rule in `Meta/RESOLVER.md` applies to a natural-language question. The skill itself does NOT call an LLM. It reads RESOLVER.md, parses every rule row, and returns either a decisive match, a ranked candidate list, or "no rule matches." The host Claude session does the natural-language understanding on top of the structured output. ## When to run - An operator wonders which rule governs a recurring question (pricing, deploy, refund, hiring exception). - A new teammate wants to find the policy for a scenario without reading the full resolver index. - Any time the resolver layer should answer a query and the operator wants the structured candidate set in one call. ## How it works 1. The skill reads `Meta/RESOLVER.md` from the vault root. 2. It parses the YAML frontmatter for the build timestamp and counts. 3. It parses every row of the `## Rules` table into a structured record. 4. It runs a deterministic match against the question: a. Tokenize the question (lowercase, drop stopwords, keep stems). b. For each rule, score by token overlap against `rule_id`, source path, skill link, and source-file H1/topic when available. c. If exactly one rule scores >= the decisive threshold, return that rule directly. d. Otherwise return up to `--limit` rules ranked by score. e. If no rule scores above zero, return `no rule matches this query`. 5. The host session then reads the structured output and frames the answer to the operator. ## Step