create-domain-map

create-domain-map

Aggregate captures into a single, version-controllable document per target domain.

Where the map is written

Resolve in this order:

1. If config.json has a private_maps_repo_path set, write into <private_maps_repo_path>/maps/<domain>/.
2. Otherwise, write into ${CLAUDE_USER_DATA:-${XDG_DATA_HOME:-$HOME/.local/share}/claude-plugins}/browser-data-capture/data/maps/<domain>/.

Tell the user which root was used. Encourage routing through a private repo (via init-private-repo) if they haven't already — domain maps benefit from version control and a visible history of changes.

Inputs

• Domain name (e.g. example.com).
• Optional: one or more capture run directories to fold into the map. If not supplied, list recent runs that contain entries for this domain and ask which to include.

What to produce

Inside <root>/maps/<domain>/:

• README.md — the canonical map. Sections:
  • Target — domain, brief description (user-supplied or inferred).
  • Status — date of last capture, total endpoints catalogued, capture sources used.
  • Auth — observed auth scheme, where credentials are passed, any rotation/expiry signals.
  • Endpoints — table grouped by resource: method, path template, purpose (best-effort guess), response status codes seen, sample size.
  • Schemas — link to schemas/ subfolder for per-endpoint JSON Schemas.
  • Pagination — detected style (offset, cursor, page-token, none).
  • Rate-limit signals — Retry-After, X-RateLimit-*, observed 429s.
  • ToS / robots / API policy — short note on what the user has confirmed about whether programmatic access is permitted. Empty by default; populated by hand.
  • Integration notes — free-form section the user maintains.
  • History — append-only log: date, what changed, which capture run contributed.
• endpoints.json — merged inventory across runs.
• schemas/ — merged JSON Schemas.
• openapi.yaml — most-recent generated spec, if generate-openapi has been run.
• captures/ — symlinks (or, in the private-repo case, references) to the source capture runs that fed the map.

Method

1. Resolve <root> per the rules above.
2. If <root>/maps/<domain>/ exists, load existing endpoints.json and merge new captures into it. Otherwise create fresh.
3. For merge:
   • Same (method, host, path-template) → union the query params, sample values, status codes, header observations. Re-infer schema from the union of body samples.
   • New endpoints → add.
4. Re-render README.md from the merged inventory + the user's hand-written sections (ToS, integration notes, target description). Preserve hand-written content — never overwrite the user's prose; only refresh the auto-generated tables.
5. Append to the History section: date, capture run IDs folded in, count of new endpoints / changed schemas.
6. If the map lives in a private repo (case 1), stage the changes but do not auto-commit unless the user asks — let them review the diff first.

Privacy

The map is built from already-redacted endpoints.json files. Do not re-introduce raw secrets. If a hand-written note in README.md contains something that looks like a credential, flag it and ask before committing.