/ field notesHANDOFF / REUSE AND VERIFICATION
Start with
a real recipe.
Keep the observed shell and shared files. Extend the selected component, then verify every requirement of the change.
Take the starting files
The archive contains fourteen page recipes, the component catalog, local fonts, exact source assets and an extension contract. Recorded vector-review screenshot crops remain in this online guide. The starter runs on any static server.
For developer work, start with the documentation recipes. Their shell, typography, navigation and code treatments differ from the marketing pages.
Assessment status
The tenth trial passes all nineteen functional criteria, with twenty-one Independent and three Assisted stages. One of four tasks passes every handoff stage. The complete handoff gate remains unmet.
The first three implementers selected four tasks from the independent product inventory. The fourth trial used a separate task selector and a sealed task set before implementation edits. Later trials retain that task set when testing a revised reference. The six stages are find, choose, specify, implement, verify and explain. Original reports and subsequent verification remain available.
Retrieve current trial dispositionsJunior Round10
Separate verification passes all 19 functional criteria. Twenty-one stages are Independent and three are Assisted; one of four tasks passes every stage. The implementer did not run all named tab-key, lifecycle, safety, retry-context and offline checks before handing off. The independent verifier ran those checks successfully, which establishes behavior but does not retroactively make the implementer's verification independent. All input/output manifests, seven changed files and the 389-file reconstruction remain intact. This context-limited trial requested Sol high and does not establish default-model or human junior usability. The current extension guidance was clarified after this sealed trial; that wording has not had a new handoff trial.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: The existing Quick start request now has Default request and Retry policy tabs. Generic tab sources use inert template text and preserve the existing install-tab fallback.; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
| new-page-or-feature | Create a public guide for handling a failed API request | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: The cloned cookbook shell contains the request, selectable success and documented rate-limit error, recovery instructions, local state controls and contract. Cookbook navigation and docs search discover the route.; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: The exported wrapper renders caller-owned request text and delegates finite outcome selection, async loading, retry and cleanup to mountFiniteOutcome. The page mounts interactive and empty instances.; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: Fixed-delay injected loaders force success, empty, error and an overlapping slow-success then fast-error sequence. The responsive CSS stacks request and outcome below 800px.; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
Junior Round9
Failed complete-task verification. Seventeen of nineteen acceptance criteria passed. Twelve stages were Independent and twelve Blocked; none of the four tasks passed every stage. Three tasks met their functional criteria. The reusable pattern still owns hard-coded data and has no caller API or executable empty-collection branch. The original verification replay exits 0 but substitutes prose checks for those behaviors and omits required checks. Source-location records are incomplete, and timestamp fields that predate the trial do not establish contemporaneous planning. The frozen input and output seals remain intact. This trial requested Terra high instead of the default Luna medium; the model escalation does not replace earlier evidence.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: Recorded; not self-rated; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
| new-page-or-feature | Create a public guide for handling a failed API request | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: Recorded; not self-rated; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: Recorded; not self-rated; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | See report | find: Recorded; not self-rated; choose: Recorded; not self-rated; specify: Recorded; not self-rated; implement: Recorded; not self-rated; verify: Recorded; not self-rated; explain: Recorded; not self-rated |
Junior Round8
Failed complete-task verification. Twelve of 19 acceptance criteria passed; 8 stages were Independent, 15 Assisted and 1 Blocked, with 0 of 4 complete tasks. The new page lacked incoming discovery and a complete docs shell. It used an undocumented exception name, ignored its caller request, and omitted tab-panel relationships. The empty fixture used success semantics. Unrelated tabs also triggered an inherited keyboard-handler error. The implementer's 19-check replay exited 0 but omitted material requirements, including the retry page. Every input, archive, reconstruction, changed-file and output seal remained intact. The verifier disclosed historic summary text in a tool response; its exact source could not be established, weakening clean-context isolation. Original observations and the provenance correction are retained.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | See report | find: Inspected examples/quickstart.html and assets/docs-recipes.js for docs-code-block/docs-code-tabs.; choose: Use a bounded retry-policy tab with no execution.; specify: Retry tab changes the local code panel and labels InvalidRequestError handling.; implement: examples/retry-policy.html and docs-recipes.js retry branch.; verify: 19-assertion Playwright run covers tab and recovery behavior.; explain: Retry only transient failures; client errors require correction. |
| new-page-or-feature | Create a public guide for handling a failed API request | See report | find: Inspected docs shell, sidebar, quickstart and cookbook recipes.; choose: Add examples/request-outcomes.html in the docs shell.; specify: One request, success, empty result and InvalidRequestError fixture with recovery.; implement: examples/request-outcomes.html.; verify: Playwright checks shell, fixtures, no API requests and narrow layout.; explain: All content is deterministic local documentation. |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | See report | find: No request-to-outcome component appeared in the recipe inventory; existing docs-code-block supplied styling conventions.; choose: Create renderRequestOutcomeExample with finite outcomes[] contract.; specify: Caller supplies request/outcomes; component renders selected result, error or empty state.; implement: assets/request-outcomes.js and assets/request-outcomes.css.; verify: Playwright checks selected state, keyboard, focus, copyable pre and empty label.; explain: No fetch or credential coupling; first use is the guide page. |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | See report | find: Reference docs use local tabs and live status conventions.; choose: Use tokenized local setTimeout fixtures with stale-token guard.; specify: Loading, success, empty, error, rapid selection, responsive and live announcements.; implement: assets/request-outcomes.js/css and guide markup.; verify: Desktop/narrow screenshots and rapid-selection assertions pass.; explain: Latest selection wins and no external side effect occurs. |
Junior Round7
Failed complete-task verification in a fresh independent context. Sixteen of 19 acceptance criteria passed and 8 of 24 stages were Independent; all four complete tasks failed. The new page lacked an existing discovery link, the outcome pattern was a hard-coded singleton whose second instance changed the first, and an Empty option did not handle an empty outcome collection. The retry variant and all five cross-cutting behavior criteria passed their observable checks, but the component anatomy and import were incomplete. The implementer's ten-check script passed on reexecution while omitting material acceptance evidence, so all Verify stages remained Blocked. Both frozen and submitted manifests remained intact.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-page-or-feature | Create a public guide for handling a failed API request | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
Junior Round6
Failed complete-task verification. Fifteen of 19 acceptance criteria passed and 11 of 24 stages were Independent; none of the four tasks passed every required stage. The submission used the supplied callable component API, but did not extend the existing code example in context, retained an incomplete shell, guessed a request/result shape and exposed a light-theme contrast defect. The cross-cutting task passed all five behavior criteria but lacked complete Verify and Explain evidence. The verifier runtime retained prior-round context; all findings were re-derived from this sealed output and fresh tests. Input and output manifests remained intact.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-page-or-feature | Create a public guide for handling a failed API request | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
Junior Round5
Failed on independent verification. All four tasks failed; 11 of 19 acceptance criteria passed and 5 of 24 stages were Independent. The retry tab lacked keyboard access, the new page had broken search and TOC links plus unsupported API details, and the outcome pattern lacked a caller API and valid panel relationships. Local state, race and narrow-layout behavior passed its observable checks. Frozen input and sealed output hashes remained intact.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-page-or-feature | Create a public guide for handling a failed API request | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | See report | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
Junior Round4
Failed on independent verification. All four tasks failed; 11 of 19 acceptance criteria passed and 6 of 24 stage claims passed. The implementation did not restore default code, omitted the complete docs shell and a reusable API, clipped guidance on narrow screens and provided no visible loading state. The task seal predates implementation edits, but conflicts with the implementer's reported start time.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Add a retry-policy variant to a high-use documentation code example | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-page-or-feature | Create a public guide for handling a failed API request | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-reusable-component-or-pattern | Introduce a reusable request-outcome example pattern | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| difficult-cross-cutting-case | Make local request-outcome examples correct across transient and narrow-screen states | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
Junior Round3
Failed on independent verification. The implementation dropped navigation and API details, did not fulfill the claimed copy fallback or all transient states, and recorded task selection after implementation had begun. The reusable factory and several behavior checks passed, but none of the four full tasks passed.
The implementer reported 18 passing checks and 0 failing checks. These counts apply only to the recorded checks.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Extend the AsyncTypeSafeClient API recipe with a usable response-code copy control and retained docs shell behavior | Independent Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-page-or-feature | Add a local SDK request explorer for score, choice and noul request fixtures | Independent Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-reusable-component-or-pattern | Add reusable ApiResponseCard rendering with two independent instances | Independent Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| difficult-cross-cutting-case | Make the request explorer cover async, failure, offline, responsive, keyboard, accessibility, safety and feature-flag behavior | Independent Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
Junior Round2
Failed on independent verification. All four task passes were rejected: 5 claimed stages passed verification, 10 failed and 9 remained unverified. Findings include an impossible selection timestamp, inactive inherited controls, clipped responsive content, absent reusable rendering API and broken clipboard recovery.
The implementer supplied command-level verification notes. The linked report retains those claims; the independent disposition above governs task completion.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| existing-component-extension | Extend the existing documentation code example with a Copy button and copied confirmation state. | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-page-or-feature | Add a local SDK request explorer page for choosing a primitive, viewing a request, and inspecting a response. | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| new-reusable-component-or-pattern | Introduce a reusable ApiResponseCard pattern for structured SDK response, metadata, and error presentation. | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| difficult-cross-cutting-case | Make the local request explorer handle loading, error, offline, responsive transformation, accessibility, and feature-flagged response details. | Pass | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
Junior Round1
Failed on author verification. The prototype loads new standalone styling instead of extending the documentation recipe; tab arrows do not work, the named feature flag has no implementation, and prompt text is inserted as executable HTML. The implementer's six passing checks did not cover those requirements.
The implementer reported 6 passing checks and 0 failing checks. These counts apply only to the recorded checks.
| Task | Selected work | Reported status | Reported stages |
|---|---|---|---|
| T1 | Extend an API reference member row with copy and response metadata | completed | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| T2 | Build a model playground quickstart page | completed | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| T3 | Introduce a typed code sample with language tabs and copy | completed | find: Independent; choose: Independent; specify: Independent; implement: Independent; verify: Independent; explain: Independent |
| T4 | Make the playground resilient across loading, offline, concurrency, responsive and accessibility states | completed-with-gap | find: Independent; choose: Independent; specify: Assisted; implement: Independent; verify: Independent; explain: Independent |