Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Execute Flexport secondary workflow: commercial invoices, products catalog, and freight invoices. Use when managing commercial invoices for customs, maintaining product catalogs, or handling freight billing through the Flexport API. Trigger: "flexport invoice", "flexport products", "flexport customs documents".
.claude/skills/jeremylongshore-flexport-core-workflow-b/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 68% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 45% | 0% |
Treat structured trade records and uploaded files as separate governed operations. Validate referenced network entities first, apply only documented create/update behavior, and never assume an uploaded document can be updated in place.
Choose purchase order, commercial invoice, or document from the business outcome; do not overload one endpoint with another resource's fields.
Verify network entity, location, shipment, and product identifiers with read-only lookups before mutation.
Confirm the OAuth credential includes the required endpoint resource. Commercial invoice create/update also requires Flexport-side enablement and permission.
Persist an operation digest, then use the documented endpoint: /purchase_orders, /commercial_invoices, or /documents. Base64 document content must remain within the documented 10 MB pre-encoding limit.
Store opaque returned IDs and retrieve the resource. For documents, replace through an approved new upload rather than inventing an update call.
Link the Flexport resource to the approved source revision, actor, version header, and redacted outcome.
REST calls authenticate with a cached OAuth 2.0 client-credentials Bearer token using audience https://api.flexport.com, or an explicitly accepted broad API key. Use distinct credentials per workload and never log credentials or tokens. MCP calls use the authenticated connection to https://mcp.flexport.com/mcp and remain subject to each tool's documented account permissions.
Use Read and Grep for discovery and evidence. Use Write or Edit only for the approved artifact, code, configuration, test, or receipt described by this workflow; do not make an unapproved Flexport-side change.
Return a machine-reviewable receipt in this shape; adapt the operation values, but never place credentials or provider payloads in it:
yamlsurface: rest-v3 operation: shipment-read decision: approved outcome: verified evidence: release_sha: recorded-out-of-band provider_reference: redacted rollback_owner: logistics-platform
A validated commercial invoice is enabled for the client, submitted once, read back by its opaque ID, and linked to the source revision. A corrected file becomes a separately approved document rather than an undocumented in-place edit.
| Failure | Response | | --- | --- | | 422 response | Correct the source payload; do not retry unchanged validation failures. | | Commercial invoice disabled | Stop and request Flexport enablement/permission. | | Document too large | Reject before base64 encoding and route to an approved alternative. | | Ambiguous create | Retrieve or reconcile by known business references before resubmission. |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 9,694 | 12,645 | +30% | 1 | 1 | 0% | 1,984 | 3,107 | +57% | 0 | 0 | — |
case-02 | fail→pass | 17,010 | 9,912 | -42% | 1 | 1 | 0% | 2,631 | 3,663 | +39% | 0 | 0 | — |
case-03 | fail→pass | 20,455 | 13,382 | -35% | 1 | 1 | 0% | 3,253 | 3,256 | +0% | 0 | 0 | — |
case-04 | pass→pass | 13,857 | 4,845 | -65% | 1 | 1 | 0% | 1,814 | 2,384 | +31% | 0 | 0 | — |
case-05 | fail→pass | 13,286 | 11,500 | -13% | 1 | 1 | 0% | 1,702 | 2,852 | +68% | 0 | 0 | — |
case-06 | fail→pass | 17,524 | 9,217 | -47% | 1 | 1 | 0% | 2,314 | 3,362 | +45% | 0 | 0 | — |
case-07 | pass→pass | 14,299 | 10,600 | -26% | 1 | 1 | 0% | 1,599 | 2,535 | +59% | 0 | 0 | — |
case-08 | pass→pass | 9,972 | 4,400 | -56% | 1 | 1 | 0% | 1,429 | 2,338 | +64% | 0 | 0 | — |
case-09 | pass→pass | 9,662 | 9,554 | -1% | 1 | 1 | 0% | 1,653 | 2,165 | +31% | 0 | 0 | — |
case-10 | pass→pass | 18,762 | 9,839 | -48% | 1 | 1 | 0% | 2,477 | 2,331 | -6% | 0 | 0 | — |
case-11 | fail→pass | 14,552 | 2,120 | -85% | 1 | 1 | 0% | 1,889 | 1,766 | -7% | 0 | 0 | — |
case-12 | pass→pass | 14,968 | 9,102 | -39% | 1 | 1 | 0% | 1,814 | 2,280 | +26% | 0 | 0 | — |
case-13 | pass→pass | 11,579 | 7,900 | -32% | 1 | 1 | 0% | 2,306 | 2,974 | +29% | 0 | 0 | — |
case-14 | pass→pass | 12,217 | 4,241 | -65% | 1 | 1 | 0% | 2,398 | 2,169 | -10% | 0 | 0 | — |
case-15 | fail→pass | 9,381 | 9,097 | -3% | 1 | 1 | 0% | 1,619 | 2,178 | +35% | 0 | 0 | — |
case-16 | fail→pass | 12,834 | 2,522 | -80% | 1 | 1 | 0% | 1,196 | 1,787 | +49% | 0 | 0 | — |
case-17 | pass→pass | 14,194 | 3,848 | -73% | 1 | 1 | 0% | 1,813 | 2,182 | +20% | 0 | 0 | — |
case-18 | fail→pass | 7,340 | 9,041 | +23% | 1 | 1 | 0% | 1,473 | 2,224 | +51% | 0 | 0 | — |
case-19 | fail→pass | 9,424 | 7,753 | -18% | 1 | 1 | 0% | 1,764 | 1,936 | +10% | 0 | 0 | — |
case-20 | pass→pass | 17,276 | 12,896 | -25% | 1 | 1 | 0% | 2,301 | 2,926 | +27% | 0 | 0 | — |
case-21 | pass→pass | 15,136 | 14,276 | -6% | 1 | 1 | 0% | 2,820 | 3,267 | +16% | 0 | 0 | — |
case-22 | pass→pass | 11,452 | 16,355 | +43% | 1 | 1 | 0% | 2,085 | 3,456 | +66% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.