build-fix-chartlisted

# Build: /fix-chart and /fix-explore You are about to generate a `/fix-chart` skill (and optionally `/fix-explore`) customised to this company's stack. Do not generate anything until you have completed all three phases. ## Phase 1 — Inspect (read before you ask) Never ask a question the repository can already answer. Investigate, in order: 1. **Find the dbt project(s).** Look for `dbt_project.yml` (there may be several — monorepo vs multi-repo matters enormously for the generated skill). Record: project names, model directory structure, whether `schema.yml` files sit beside models or in a central folder. 2. **Identify the semantic layer.** Look for: - Lightdash: `lightdash.config.yml`, `meta` blocks in dbt `schema.yml` files - Looker: `*.model.lkml`, `*.view.lkml` files - Cube, Metabase, or metrics-layer configs 3. **Read the conventions from the code, not the docs.** Sample 5–10 model files and record: naming pattern (`fct_`/`dim_`/`stg_`?), how joins are declared, whether tests exist (`unique`, `not_null` on primary keys — their absence predicts fanout bugs), how metrics/measures are defined. 4. **Check CI and review gates.** `.github/workflows/`, `CODEOWNERS`, PR templates. The generated skill must produce PRs that pass these gates on the first attempt. Summarise findings in 5 lines before moving on. If you cannot find a dbt project, stop and say so — this builder is for dbt-backed BI stacks. ## Phase 2 — Interview (ask only what the code cannot tell you) A