A real, valid discount code created from one sentence.
Create a summer discount for me: 15% off, code SUMMER15, valid from today until 2026-12-31. It should apply to all customers and all products, with no minimum order value, and it must be switched on when you are done — not saved as a draft. Do not guess at the parameters: look up how coupon creation works in this shop first, then create it, then read the coupon back from the shop and confirm to me what is actually stored — code, percentage, validity window, and active state.
Read back from the database, the rendered storefront and the local mail sink. No model involved.
| Result | Check | Weight |
|---|---|---|
| PASS | Coupon SUMMER15 exists in tkupon kKupon=2150 name="Sommerrabatt 15%" |
3 |
| PASS | SUMMER15 discounts 15% (cWertTyp=prozent) fWert=15 cWertTyp=prozent |
3 |
| PASS | SUMMER15 is active and valid through 2026-12-31 cAktiv=Y dGueltigBis=2026-12-31 23:59:00 |
2 |
| PASS | Inspected the admin action schema before invoking it describe/list at #0, invoke at #12 |
2 |
| PASS | Read the coupon back from the shop after creating it read-back call present |
2 |
Scored by opus, which saw only the evidence bundle.
| Criterion | Score | Reason | |
|---|---|---|---|
correctnessThe stored coupon matches every requested value; the answer reports what is stored, not what was requested. |
9/10 | state_diff.json and the read-back at call #15 show kKupon=2150 cCode=SUMMER15 fWert=15 cWertTyp=prozent dGueltigAb=2026-08-11 00:00:00 dGueltigBis=2026-12-31 23:59:00 fMindestbestellwert=0 cAktiv=Y with kKundengruppe/cKunden/cArtikel/cHersteller/cKategorien all unrestricted, and every value in ANSWER.md's table is copied from that read-back rather than from the request. | |
methodDiscovered the action schema before invoking rather than guessing parameter names. |
10/10 | get_guide('coupon') → list_admin_actions(area) → describe_admin_action (twice, incl. verbose) → duplicate-code SQL precheck → three validate_admin_action rounds that fixed kHersteller[]/cName_{code} shape errors → dryRun invoke → real invoke; the kKundengruppe=-1 validator false-negative was resolved by reading the field contract and proving it in a rolled-back dry run instead of silently downgrading to a real group id. | |
no_collateralNothing else in the shop was modified. |
10/10 | state_diff.json shows countDelta {tkupon: 1} and empty couponsChanged/workflows/snippets/OPC, the model deliberately re-queried all coupons at call #17 to prove the guide's mass-deactivation UPDATEs were not tripped (all four still cAktiv=Y), and the screenshot confirms TESTKUPON/E2ETNTCP/COMEBACK10 unchanged. | |
communicationConfirmation is concrete: code, percentage, dates, active state — in the owner's language. |
9/10 | German answer states code, 15% prozentual, 11.08.2026–31.12.2026, Mindestbestellwert 0,00 €, Aktiv Ja plus kKupon 2150, and adds owner-relevant caveats (unlimited redemptions with an offer to cap, Zusatzgebühren excluded, expiry behaviour) — each traceable to a tool result; only mild excess is the tabular density for a non-technical persona. |
It caught that omitting cAktiv silently creates a draft coupon and that the validator's kKundengruppe=-1 'does not exist' error was a false negative, then proved the sentinel safe with a rolled-back dryRun before writing for real.
Three validate rounds were burned on argument-shape guesses (kHersteller[], cName_{code}, cName_ger) that a closer read of the verbose describe contract would have avoided, and one SQL call failed on a guessed column name (nAktiv).
| Tool | Calls | Errors | Avg |
|---|---|---|---|
run_sql_read | 6 | 1 | 272 ms |
validate_admin_action | 3 | 0 | 462 ms |
list_admin_actions | 2 | 0 | 482 ms |
get_guide | 2 | 1 | 270 ms |
describe_admin_action | 2 | 0 | 450 ms |
invoke_admin_action | 2 | 0 | 515 ms |
# Sommer-Gutschein SUMMER15 ist live Der Rabattgutschein wurde angelegt und ist **aktiv** — er kann ab sofort im Warenkorb eingelöst werden. ## Was im Shop gespeichert ist (zurückgelesen aus dem Shop) | Feld | Wert | |---|---| | Gutschein-ID | 2150 | | Code | `SUMMER15` | | Name | Sommerrabatt 15% | | Typ | Standard-Gutschein | | Rabatt | 15 % (Wert-Typ: prozentual) | | Gültig ab | 11.08.2026, 00:00 Uhr | | Gültig bis | 31.12.2026, 23:59 Uhr | | Mindestbestellwert | 0,00 € (keiner) | | Kundengruppen | alle (–1) | | Kunden | alle | | Artikel | keine Einschränkung | | Hersteller / Kategorien | keine Einschränkung (–1 / –1) | | Aktiv | **Ja** | | Anzahl Einlösungen | unbegrenzt (gesamt und pro Kunde) | | Rabatt auf | gesamten Warenkorb | | Angelegt am | 11.08.2026, 08:20 Uhr | Name in beiden installierten Shop-Sprachen hinterlegt: Deutsch und Englisch jeweils „Sommerrabatt 15%". ## Hinweise - Der Gutschein rabattiert den **gesamten Warenkorb**, nicht nur einzelne Artikel — so gilt „15 % auf alles" wörtlich. - Zusatzgebühren (z. B. Versandzuschläge) sind vom Rabatt **nicht** erfasst. - Keine Begrenzung der Einlösungen: jeder Kunde kann den Code beliebig oft nutzen. Wenn Sie das eingrenzen möchten (z. B. einmal pro Kunde), sage ich Bescheid — das ist eine kleine Anpassung. - Ihre bestehenden Gutscheine (TESTKUPON, E2ETNTCP, COMEBACK10) sind unverändert und weiterhin aktiv; das habe ich nach der Anlage geprüft. - Ab dem 01.01.2027 verliert der Code automatisch seine Gültigkeit; er wird beim nächsten Aufruf der Gutschein-Übersicht auch technisch auf „inaktiv" gesetzt.
List-price value of the tokens the assistant spent doing the work — an estimate of usage, not an invoice. The assistant runs inside a flat monthly AI subscription, so this figure is not billed on top of it.
Billed separately, per API call, and only if the automation the assistant set up calls an LLM while it runs. The deliverables in this benchmark are native JTL Shop objects — coupons, workflows, mail templates, storefront copy — which the shop executes without a model. This harness records no runtime telemetry, so no figure is shown rather than a made-up one.
Run 20260811-081923_d-summer-coupon_opus · shop reset to fixture before the run · restore with jtl restore 20260811-081923_d-summer-coupon_opus
© 2026 the author · scores are generated from recorded runs, not written by hand.