Product imports fail and nobody knows which rows to fix

The problem

A product launch stalls when supplier spreadsheets contain invalid headers, duplicate handles or broken variant groupings.

Agent job

Write a schema-aware preflight checker; identify risky overwrites and variant collisions; propose a corrected COPY with a row-by-row change log. Never infer missing identifiers.

Expected output

Corrected CSV, exception workbook, reversible import instructions and checker script.

Why an agent / simpler alternative

Assessment

The agent can reconcile a supplier’s unusual column names with the merchant’s documented schema and explain ambiguous variant relationships. Deterministic checks should enforce encoding, required columns and uniqueness. Use the built-in importer or a simple script alone when the mapping is already stable; manually resolve product identity.

Setup requirements

Items

1. Read-only catalog export, error log and dated platform import specification

2. Separate input/output directories and backups; never work on the only copy

3. CSV parser and a checker whose rules can be inspected

4. Merchant-selected examples of valid multi-variant products

Inputs and access

1. Name

   Merchant-owned product CSV

   Source or owner

   Prospective consenting customer or specified official publisher

   Sensitivity

   Use synthetic, public or appropriately deidentified material; confirm actual sensitivity before work

   Permission needed

   Owner authorizes scoped access and processing; verify rights and retention before receiving files

2. Name

   Current exported catalog and desired changes

   Source or owner

   Prospective consenting customer or specified official publisher

   Sensitivity

   Use synthetic, public or appropriately deidentified material; confirm actual sensitivity before work

   Permission needed

   Owner authorizes scoped access and processing; verify rights and retention before receiving files

3. Name

   Store schema and import error log

   Source or owner

   Prospective consenting customer or specified official publisher

   Sensitivity

   Use synthetic, public or appropriately deidentified material; confirm actual sensitivity before work

   Permission needed

   Owner authorizes scoped access and processing; verify rights and retention before receiving files

Proposed steps

1. 1

   Freeze the original export and record file hashes, row counts and expected product/variant counts.

   Expected output: Input manifest and untouched backup

   Approval gate: Owner confirms catalog scope; do not pull a full customer/order export.

2. 2

   Compare supplied headers to the current import schema; draft a mapping with confidence and unanswered questions.

   Expected output: Column mapping and exception queue

   Approval gate: Merchant resolves uncertain identifiers, options and inventory meanings.

3. 3

   Run inspectable deterministic checks for encoding, required values, duplicate identities and variant structure.

   Expected output: Row-indexed validation log

   Approval gate: Do not silently merge rows or invent missing GTINs.

4. 4

   Generate a corrected copy and machine-readable before/after diff; enumerate possible live overwrites.

   Expected output: Candidate CSV and change ledger

   Approval gate: No store upload; reject changes beyond approved fields.

5. 5

   Run synthetic seeded tests and manually reconcile unaffected SKUs and sampled variant families.

   Expected output: Acceptance report with preserved totals

   Approval gate: Critical missed errors or uncertain variants block delivery as import-ready.

6. 6

   Hand over the copy, unresolved cases and staging/rollback instructions; define who accepts the import result.

   Expected output: Reviewed delivery pack

   Approval gate: Merchant approves exact overwrite behavior and separately authorizes any import.

Synthetic output illustration — not an executed result

Synthetic

Yes

Executed

No

Text

SYNTHETIC: Row 18, SKU BLUE-M: proposed header change from Colour to Option1 Value. Row 22: BLOCKED, two different products share handle blue-shirt. Owner decision required. Unchanged SKU count: 28/28. This is an invented output illustration, not a test result.

Human approval boundaries

1. Merchant approves mapping and every overwrite; human performs or separately authorizes the backed-up staging import.

2. Human service provider reviews quality, scope and quoted fee; customer approves contract, delivery recipients and payment terms. Separately approve purchases, credentials, publication and external sends. No commitments or payments are made in this research.

Failure modes

1. Unique handles may legitimately repeat across variants

2. Wrong inventory assumptions can overwrite live records

3. Image URLs and encoding can fail after local validation

Recovery and troubleshooting

Items

1. If a variant relationship is ambiguous, quarantine its whole product family rather than repairing individual rows.

2. If a checker misses a seeded defect, add a deterministic regression check and rerun the full fixture.

3. If an authorized staging import fails, retain its error log, restore the staging backup if needed and correct only the documented issue; do not retry in production.

Unexecuted validation plan

Sample setup

Create 30 synthetic products with 10 seeded header, variant and encoding errors; run on a copy.

Acceptance checks

1. Detect all seeded high-risk errors

2. Preserve every unaffected SKU and variant relationship

3. Human verifies 10 corrected rows

4. No production import in the pilot

Stop conditions

1. Stop if data authority or rights are unclear.

2. Stop if critical seeded failures are missed; do not claim readiness.

3. Stop if manual correction time exceeds the pilot effort budget without customer approval.

Executed

No

Costs and effort — unknown or hypothetical

Status

unknown_or_estimate

Currency

CAD

Items

1. Agent subscription/API usage: not quoted; verify current plan and commercial data terms.

2. Human intake, rights review, QA and revisions: unknown until timed pilot.

3. Storage, hosting, integration and payment-processing costs: excluded unless separately agreed.

4. Customer acquisition and support time: not measured.

Caveat

No margin or income estimate. Price hypotheses are editorial offer tests, not observed market rates; taxes, scope and effort require a real quote.

Proposed service concept — earnings not verified or guaranteed

Buyer

Small Shopify merchant or ecommerce implementation agency

Deliverable

Corrected CSV, exception workbook, reversible import instructions and checker script.

Charging basis

One-off catalog rescue package; future batches separately scoped.

Acquisition channel

Permission-based discovery with existing professional contacts, referrals or an opt-in sample page; none executed.

Assumptions

1. Customer has this unresolved problem and values a reviewed deliverable above existing built-in alternatives.

2. A consenting pilot confirms data access, achievable quality and economically tolerable review time.

3. Provider can deliver within a written scope and supports the result.

Revenue guarantee

No

Canadian adaptation — proposed

Proposed Canadian pilot: preserve merchant-approved CAD values, bilingual product text, accents and metric/imperial units without conversion guesses; test the actual store export schema and UTF-8 import. Confirm Canadian-store account access and backup permissions. No Canadian deployment or compliance claim.

Limits

1. Original proposed plan; untested.

2. Problem and capability references are evidence of context only, not demonstrated end-to-end success, customer demand, willingness to pay or profitability.

3. Compatibility, workload, delivery effort and availability must be verified in a real authorized pilot.

4. Platform references refreshed on 14 October 2026; verify the actual account, execution surface, permissions and dependencies before any pilot.

Candidate platforms — compatibility untested

1. Name

   Codex

   Required capabilities

   1. File analysis, code execution and test/report artifact creation.

   Capability source urls

   1. https://learn.chatgpt.com/docs/codex/cli

   2. https://learn.chatgpt.com/docs/artifacts-viewer

   3. https://learn.chatgpt.com/docs/sandboxing

   Workflow compatibility

   Unknown; verify in the proposed first test.

   Historical candidate name

   Codex

   Selection basis

   Editorial candidate from general capabilities; exact end-to-end fit untested.