What the documents look like
Cross-border product compliance happens before listing and import, and its central evidence concerns product safety and labels. The documents handled day to day include product test reports, English labels, MSDS files and platform prohibited-item lists. Sources include platform policy centres, testing laboratory reports, destination-country regulation libraries and supplier materials.
This page covers the task of product and marketing copy drafting. Whether a promotional sentence in a draft is supported by the materials depends on what the source text records about the subject, the object, the status and the valid scope. The role of workflow routing here is to place the draft copy and the corresponding materials into a structure that can be checked, to collect the information the draft requires, and to let a user selection determine the following branch. It does not replace the compliance judgement itself.
Two facts shape this step. First, the materials differ in what they record: a test report, a label draft and a supplier statement each carry different fields, and a draft sentence can be checked only against the fields the material actually states. Second, the update rhythm of these materials is irregular and tied to each new listing, so a routing design cannot rely on a fixed refresh pattern; it must work from the material text supplied at the time of drafting.
Constraints at the workflow routing step
The form input node is a user interaction node. When it is triggered, the conversation enters an interaction state, the workflow state is recorded, and after the user completes the interaction the workflow continues. Once the user fills in the required information and submits, the node collects the submitted form information and passes it to subsequent nodes. The user selection node is also a user interaction node, with the same interaction-state behaviour. When the user makes a choice, the node judges that choice and executes the corresponding "yes" branch. Both nodes can be added repeatedly and can trigger execution.
Information collection and branch selection are existing node behaviours. Order writes, document backfill and external status synchronisation require a separate business interface or a subsequent process.
Hypothetical paper exercise sample (paper materials, not real records):
Material text one: "The inner liner of this stainless steel vacuum flask is made of 304 stainless steel. Test report number LAB-2024-0317, report date 17 March 2024. The test object is heavy metal migration from the inner liner."
Material text two: "The English label draft states a capacity of 500ml and marks 'BPA Free'. Label version date 2 April 2024. Applicable market: Germany."
Material text three: "The supplier material states that this batch of product does not contain bisphenol A. Material issue date 20 February 2024. No applicable market is stated."
Draft sentence to be assessed: "This product has passed German food-contact material testing and is suitable for infant use."
The subject of this draft sentence is the vacuum flask; the objects are the German market and the infant-use scenario. Material one records only heavy metal migration testing of the inner liner. Material two records only label marking content. Material three states no applicable market. The claims "has passed German food-contact material testing" and "is suitable for infant use" have no corresponding record in either material. The assessment result is unsupported, and the evidence gap is retained.
Configuration and implementation checks
| Configuration item | Hypothetical starting value for this exercise | How to check |
|---|---|---|
| Form input node fields | Draft sentence text, material source text, target market | Compare each submitted value with the value read by subsequent nodes |
| User selection node branch | Two-way choice: "does the material cover the subject and object of the draft sentence" | Check the selection result against the connected "yes" branch that executed |
| Interaction state record | Conversation enters interaction state after the node triggers | Observe whether the dialogue is hidden and whether the state is recorded |
| Downstream receiving node | Pass collected information to a checking node | Compare output field names with input field names |
| External operation division | Final copy backfill connects to a separate interface | Check whether this is declared as a subsequent process |
The form input node field starting values are the draft sentence, the material source text and the target market. The reason for choosing these three is that the assessment in this sample requires all three to compare subject and object. Compare them with the fields the downstream node actually reads; a missing field is recorded as a gap. This starting value is a hypothetical configuration for this exercise and is not an official default.
Three common errors
- Treating material three's "does not contain bisphenol A" as support for the draft sentence's "suitable for infant use". Input: material three's source text and the draft sentence. Comparison: whether material three records infant applicability. Observable result: material three states no applicable market and no applicable population, so the assessment is unsupported.
- Treating material two's label marking "BPA Free" as evidence that German food-contact material testing was passed. Input: material two's source text and the draft sentence. Comparison: the label marking content against a test conclusion. Observable result: material two records no test conclusion, so the assessment is unsupported.
- Treating the default execution of the user selection node's "yes" branch as confirmed material coverage. Input: the user selection result and the branch connection. Comparison: the selected value against the branch actually executed. Observable result: when the branch is not connected or the selection is not submitted, the downstream node should not read a confirmation value.
How to verify the setup
- Enter material texts one, two and three together with the draft sentence into the form input node. After submission, check the fields read by the subsequent node. Compare the read fields with the submitted fields; if they match, collection and transfer hold. If a field is missing, record the gap.
- In the user selection node, select "the material does not cover the subject and object of the draft sentence" and check the "yes" branch that executes. Compare the selected value with the branch connection; if the corresponding branch executes, the judgement holds. If it does not execute, record a connection problem.
- Use a paper input of an empty result to check the downstream design. Set the retrieval result to an empty list and check whether the subsequent node handles it under an empty-result branch. Observe whether the downstream declares empty-result handling; if it does not, record this as pending implementation. Do not claim that a fictional rule retrieval necessarily returns an empty list.
A relevance score is not a category, entity or permission filter, and retrieval is not authorisation. Same-model retrieval association applies to the documented condition; it does not extend to unrelated external systems. Forms collect values; business verification, approval, database writes and external synchronisation require their stated downstream implementation.