Rp Discovery

Inputs

Expected inputs may include:

• migrations/<project>/orchestration/run.json
• migrations/<project>/orchestration/decisions.json
• source site/app URL
• source platform name when it cannot be inferred
• source acquisition mode when the platform offers multiple read paths
• source API docs
• credentials, tokens, or local dump files when available
• export files when the run is file-based rather than URL-based
• current project under migrations/<project>/
• project-local config under migrations/<project>/config/

Config gate before capture

Before running source capture, verify the project-local config files created by replatform:

• config/wix.env should always exist with WIX_SITE_STRATEGY, WIX_SITE_ID, and WIX_AUTH_TOKEN keys, even though discovery itself may not use Wix credentials yet.
• config/source.<platform>.env should exist once the source platform is known. For WordPress this is config/source.wordpress.env.

If a required key is missing or blank, ask the user for that value and fill the config file for them before continuing. Ask one value at a time. Never print secret values back to the user; report only present/missing.

Treat migrations/<project>/config/*.env as secret-bearing once they may contain real values. Do not inspect them with whole-file reads that echo contents into tool output. Use secret-safe checks only: existence, required key names, and present / blank / missing status.

Site creation precedence

If this skill encounters conflicting Wix guidance about how to create a new site + headless destination, replatform‘s migration contract wins.

• Route headless destination creation back to the Wix CLI headless scaffold defined in replatform → “Headless site creation” (npm create @wix/new@latest headless). This is the verified way to get a genuine headless site; the account-level Projects API is deprecated for this workflow (it produced non-headless sites).
• Discovery is source-side and does not create the site itself — just defer to that section.

Workflow

1. Confirm the active project under migrations/<project>/.

2. Start from the source URL when available and try to identify the source platform yourself before asking the user. Use lightweight signals such as a REST index, headers, HTML/application markers, or platform-specific route patterns. Only ask the user to name the platform if detection remains inconclusive.

3. Once the platform is inferred, resolve the acquisition mode before requesting source credentials when the platform has materially different read paths.

   • For Shopify URL-based migrations, ask whether to use the Admin API or only publicly available storefront data.
   • For WordPress / WooCommerce URL-based migrations, ask whether to import public content only or also include private/authenticated data.
   • Only the WordPress / WooCommerce also include private/authenticated data branch should trigger a credentials request. The public content only branch proceeds without credentials and should be described as limited to public data. For WooCommerce, this branch should still probe public Store API catalog routes such as /wc/store/v1/products and /wc/store/v1/products/categories before declaring commerce out of scope. Treat file/export ingestion as a separate flow that starts from user-provided files instead of a site URL probe; do not offer exports as a third option in the URL-based acquisition-mode question.

4. Then select the matching source adapter skill (e.g. rp-source-wordpress for WordPress / WooCommerce, rp-source-csv when the run is file-based). If no adapter exists for the platform, capture entities manually following the same output contract.

   • File-based runs (sourceMode=files_only, sourcePlatform=csv) use rp-source-csv regardless of which system produced the files; that adapter identifies the originating vendor from the header row. There is no acquisition-mode question and no credentials request for this path.

5. Run the adapter’s capture step to produce a raw, machine-captured dump under <migrations-root>/<project>/data/<source>-discovery/. For WordPress, the capture script lives in rp-source-wordpress/scripts/ — run it from that skill directory (see rp-source-wordpress Capture section and CONVENTIONS.md). The adapter owns the capture mechanics, auth model, and platform quirks; this skill consumes its output. For long runs, pass --progress-log <path> and poll it per CONVENTIONS.md#progress-log-polling.

   • Distinguish supported entities (advertised by the source) from used entities (those with recordCount > 0). Entities advertised but empty should be flagged, not mapped as if they hold data.
   • A capture made without credentials is usually incomplete (gated entities, private fields, PII return 401/403). Do not treat an unauthenticated capture as authoritative — the adapter documents what auth a complete run requires.
   • For WordPress / WooCommerce captures, read data/wp-discovery/skipped-routes.json when present. Treat it as the canonical route-scope audit trail: skipped routes are evidence, not source entities, unless they were explicitly force-included by an audited override.
   • For CSV captures, the capture script lives in rp-source-csv/scripts/csv-discovery.js and takes the whole file set in one run (--file is repeatable) so roles and split files resolve together. Read data/csv-discovery/fileset.json — it is the canonical machine capture, and source-schema.json is synthesized from it:
     • carry sourceFiles[] (with role, vendor, partOf), vendor, dialect, drift, mappingHints, and csvInputRoot into sourceMeta, keeping file paths relative to csvInputRoot so the project stays movable;
     • give every entity an origin (file-rows | row-group | column-values) with the parameters that origin needs, and set hierarchical: true on nested derived entities so the mapper’s faithfulness-ledger rule fires;
     • surface drift.unmappedColumns as unknowns so the mapper handles them explicitly;
     • honor halt: true. An ambiguous layout, an unknown file role, conflicting split-file headers, or a near-miss vendor detection is a question for the user, not something to resolve by picking the highest-scoring candidate. The warning text names the decision to put to them.

6. Capture field-level schema details, including type, cardinality, requiredness, and example values.

7. When bundled Wix domain knowledge recognizes a source route or source entity, annotate the discovered entity with sourceMeta.candidateTargetRefs[] such as ["stores/product"]. Discovery must still record source facts only; these refs are mapper hints, not target decisions.

8. Note operational constraints such as pagination, rate limits, auth model, and incremental sync options. If the source base URL or discovered media/file URLs use localhost, 127.0.0.1, or another private-only host, record a media reachability note in source-profile.md. Localhost is fine for discovery and local source reads, but Wix Media import is URL-based and Wix servers cannot fetch the user’s localhost. This is an optional preparation step and, as far as we know today, only affects media import. State the two acceptable choices:

   • expose the source with a public HTTPS tunnel such as ngrok before live media import
   • skip/defer media import while continuing non-media entities

   Include concise ngrok setup instructions when relevant:

   bashCopy

   brew install ngrok
   ngrok config add-authtoken "<YOUR_AUTHTOKEN>"
   ngrok http 8090
   export WP_BASE_URL=https://<id>.ngrok-free.app

9. Synthesize the raw capture into the normalized artifacts below.

Artifacts to create or update

• migrations/<project>/discovery/run.json

• migrations/<project>/discovery/entities/

• migrations/<project>/discovery/warnings.json

• migrations/<project>/discovery/llm-handoff.json

• migrations/<project>/orchestration/checkpoints.json

• migrations/<project>/data/<source>-discovery/: raw machine-captured output from the source adapter. Treated as evidence, not a hand-off artifact — downstream skills reference it for traceability but do not read it wholesale.

• migrations/<project>/source-profile.md: source platform, access method, limits, auth, and operational notes. Synthesized from the raw capture. Capture the operational facts the adapter documents (auth model, pagination, rate limits) so rp-import-codegen has them without re-deriving.

• migrations/<project>/source-schema.json: machine-readable schema for entities and fields. Synthesized from the raw capture — this and source-profile.md are the canonical hand-off to rp-mapper. Include traceability pointers so the mapper can drill into a specific entity’s raw file when needed:

  • top-level rawDiscovery: relative path to the raw capture dir, e.g. data/wp-discovery/.
  • per-entity rawFile: file name within that dir, e.g. wp-v2--posts.md.
  • per-entity recordCount and inUse so consumers can distinguish supported vs. actually-used entities.
  • per-entity relations derived from the source-declared relationships in the raw capture, so relationships are evidence-backed rather than guessed. Each relation should carry an evidence pointer back to the source signal it came from.
  • For WordPress / WooCommerce, synthesize entities only from sampled backend_data and accepted backend_metadata route artifacts. Do not synthesize entities from routes listed in skipped-routes.json unless the skipped-route record has includedByOverride: true; in that case, include originalDiscoveryCategory, includedByOverride: true, and overrideReason when present in the entity sourceMeta.
  • Follow the adapter’s source-schema.example.json for the shape (e.g. rp-source-wordpress/source-schema.example.json). It is a template to follow, not a strict schema to validate against — keep the platform-agnostic core stable and push platform quirks into each entity’s open sourceMeta blob.

• Optional supporting notes under migrations/<project>/research/ if needed.

Output quality rules

• Separate confirmed facts from assumptions.
• Record per-entity volume (record counts) so downstream skills know what the site actually uses, not just what it supports.
• Preserve source-specific identifiers exactly.
• Include enough detail for downstream mapping and code generation.
• Flag unknowns explicitly instead of inventing structure.