# Peated Catalog > Catalogs or repairs Peated Bottle records directly for a named Bottle, brand, Series, distillery, bottler, or release set. Use for missing Bottles, wrong facts, Series membership, images, aliases, references, or duplicate review. Do not use for code changes or store-price queue moderation. - Skill: `dcramer/peated-catalog` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add dcramer/peated-catalog` - Raw SKILL.md: https://api.skillmd.com/api/skills/dcramer/peated-catalog/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: dcramer (https://skillmd.com/u/dcramer) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/dcramer/peated-catalog --- # Peated Catalog Finish the named catalog target in production unless the user names another environment or asks only for research or review. Follow any narrower target the user gives. ## Read what applies - Use `docs/operations/catalog-maintenance.md` for a full Bottle, brand, Series, distillery, bottler, or release-set review. - Read `docs/architecture/whisky-identity-model.md` before deciding whether Bottles are distinct, related, or duplicates. - Open only the linked guide needed for the current work. - Use `pnpm cli auth` and `pnpm cli api` for production data. Do not use legacy database commands. ## Work 1. Name the target and environment. Resolve stored IDs with read-only API calls. 2. Before deciding what to change, build a list of current and past releases from sources outside Peated. Fetch every page of Peated results and compare the lists. The producer's current range does not define the catalog. Search every period and product family, including discontinued, one-off, and country-specific releases. For a distillery scope, also search historical owner and house brands; domestic blends and older releases may not use the distillery as their Brand. 3. Use specialist catalogs, collector lists, old sites, and auction archives to find past releases. Verify each release and fact with producer pages, announcements from the time, readable labels, or exact auction records. If a search fails, try other names, languages, archives, and kinds of sources. 4. Track every release and Peated record as `create`, `update`, `merge`, `no change`, `unresolved`, or `out of scope`. Save the source for each change. Check each existing Bottle and possible missing release. Do not mark old records `unresolved` as a group because there are many or no current producer page exists. For each unresolved item, record the unanswered question and the sources or searches tried. 5. Review all fields, Series, the target Entity, images, aliases, and import references required by Catalog Maintenance. 6. Before writing, state the target and action counts. A direct catalog request allows supported creates and updates within that target. Ask before merges, deletes, uncertain identity changes, or work outside it. 7. Check the live OpenAPI schema. Re-fetch each record before changing it. Use exact IDs and send only supported fields. Stop if the record changed or the API returns a conflict or validation error. 8. Re-fetch every changed record. Check shared edits, images, aliases, references, and redirects when they apply. 9. Before reporting completion, save durable results in a dated, scope-specific artifact under `docs/research/catalog/`. Use `YYYY-MM-DD-short-name.md`, or a directory with a `README.md` when the report has supporting data. Do not append separate tasks to a shared or monthly report. For the research record, keep exact links, what each source established, the years, markets, and release families covered, material conflicts or page errors, searches tried for unresolved gaps, and image source and reuse findings when they will help later work. Distinguish leads from evidence and do not claim that a source is complete unless it proves that. Do not keep raw API replies, downloaded images, tokens, request files, or other temporary working data. Update `docs/operations/catalog-research.md` only when the task finds a reusable research method, access limit, or recurring trap. Keep task-specific results in the scoped research artifact. ## Rules - One marketed release is one Bottle. Package size or packaging alone does not create another Bottle. - Use the producer's stable product name. Keep age, year, ABV, edition, cask facts, and outturn in their fields. - Use `null` for unknown or disputed facts. Keep an existing value unless a stronger source for the same Bottle proves it wrong. - A Bottler independently selects and releases whisky made by another producer. A name on the label does not prove that role: do not infer it from an owner, importer, distributor, physical packer, customer, retailer, bar, or private barrel holder. Do not clear an existing Bottler only because a source omits it; require stronger evidence that the assignment is wrong. - Use a Series only for an evidenced, named product range owned by one Brand. Shared artwork, a client, a release year, or similar packaging does not prove Series membership. Add an alias only for a proven public name. Assign an import reference only when the full text identifies one Bottle. - Use only an image of the exact Bottle. Save the canonical page where it appears and its reuse terms, then inspect the stored image. A populated image URL, a direct asset URL, or a missing license does not by itself complete the provenance check. - Merge only proven copies of the same marketed release. Completing one family or the current range does not complete the catalog. Continue through every period and product family until each item has a final status. Report the environment, sources and years covered, counts by status, changed IDs, checks performed, research records changed, and unresolved questions.