Case ExplorerCase 03

Reconstruct a local-first Markdown workspace

Evidence and provenance

Frozen evidence (4)

evidence capture: desktop-persisted — **E01** desktop-persisted

evidence capture: desktop — **E02** desktop

evidence capture: mobile-reader — **E03** mobile-reader

evidence capture: mobile — **E04** mobile

012026-08-29official source github.com
022026-08-29live stackedit.io
032026-08-29reference-capture

Evaluation matrix

Score composition

Submission contract, install/build and policy are gates. Contract v2 preserves the full score vector instead of hiding functional checks, visual regions, responsive behavior, workflows and preservation in one value.

global 0.50 · regions 0.00 · functional 0.10 · responsive 0.10
workflow 0.30 · preservation 0.00 · floor 0.62

Active score weights are divided by their total. 1.00

Visual checkpoints (3)

reference capture for desktop — Fig. 01desktop1440×900 · viewport · motion reduce

reference capture for mobile — Fig. 02mobile390×844 · viewport · motion reduce

reference capture for mobile-reader — Fig. 03mobile-readerclick on mobile · viewport

Contract v2 evidence

1 workflow · 5 ordered steps · 0 weighted regions

Evidence	Contract	Weight
edit-persistence	replace-note fill ×1.5 → replace-title fill ×1.0 → reload navigate ×2.0 → restore-title fill ×0.3 → persisted screenshot ×0.3	2.0

Deterministic checks (6, total weight 4.5)

Check	Kind	Dimension	Target	Weight
seeded-preview	text	functional	desktop	1.0
editor-present	selector	functional	mobile	0.5
reader-control	selector	functional	mobile	0.5
document-title	selector	functional	desktop	0.5
mobile-no-overflow	no_horizontal_overflow	responsive	mobile	1.0
mobile-preview-hidden	hidden	responsive	mobile	1.0

Development diagnostics · 10 recorded attempts

10 / 10completed attempts

0infrastructure failures

0timeouts

64.5–93.1observed overall range

case idmodelmodel requestedmodel resolution statusruntimestatus

Date	Agent	Variant	Status	Visual	Func.	Overall	Time	Cost
2026-08-30	claude_code/claude-opus-5 · high	—	completed	94.8	100.0	93.1	37m 25s	$14.80
2026-08-30	claude_code/claude-sonnet-5 · high	—	completed	73.2	100.0	64.5	9m 03s	$2.43
2026-08-30	codex/unverified · high	—	completed	86.3	100.0	81.9	11m 09s	—
2026-08-30	codex/unverified · high	—	completed	83.0	100.0	77.5	6m 32s	—
2026-08-30	codex/unverified · high	—	completed	89.6	100.0	86.2	12m 50s	—
2026-08-30	claude_code/claude-opus-5 · high	—	completed	93.9	100.0	91.9	40m 06s	$15.06
2026-08-30	claude_code/claude-sonnet-5 · high	—	completed	74.4	100.0	66.0	18m 42s	$4.27
2026-08-30	codex/unverified · high	—	completed	89.8	100.0	86.5	12m 01s	—
2026-08-30	codex/unverified · high	—	completed	80.3	100.0	73.9	9m 46s	—
2026-08-30	codex/unverified · high	—	completed	90.2	100.0	86.9	15m 57s	—

Candidate protocol preview

Access-restricted Pilot

The rendered Candidate prompt is published for inspection. The downloadable starter workspace, injected Skill files and executable Harness remain restricted, so this page does not claim that an external reader can reproduce a graded run.

Skills declared design-case-implementer

Rendered Candidate prompt

Runtime-agnostic prompt record.

Use the ReconBench candidate skills `design-case-implementer` in that order. Read those skill files if your workspace provides them, and otherwise follow the same process with your own tools.

You are the candidate in ReconBench task `stackedit-field-notes` (Reconstruct a local-first Markdown workspace).

Read `task/brief.md` and every other file under `task/` before changing code.
Implement the result only under `submission/`. The starter may be replaced when the task allows it.
Do not inspect parent directories, benchmark harness files, grader code, reference images, or any oracle material.
Do not call write endpoints belonging to a reference site. Keep forms, counters, and similar behavior local.

Dependencies install offline from a vendored cache that matches the starter lockfile.
Do not add dependencies; the install will fail and the network is not yours to use.

`tools/screenshot.mjs` renders your work in the browser build the evaluator uses:

node tools/screenshot.mjs --url http://127.0.0.1:4173/ --out shots/desktop.png \
--width 1440 --height 900 --full-page

Before finishing:
1. Run the requested build command.
2. Start the app and capture every target viewport in the task with that tool, then read the captures back.
3. Exercise the required interactions and inspect the console errors and blocked requests it reports.
4. Write `submission/submission.json` using the contract in the implementer skill.

The harness will evaluate the filesystem after this process exits. Do not leave a dev server running.

Prompt rendered by the Harness for Codex CLI.

Use $design-case-implementer.

You are the candidate in ReconBench task `stackedit-field-notes` (Reconstruct a local-first Markdown workspace).

Read `task/brief.md` and every other file under `task/` before changing code.
Implement the result only under `submission/`. The starter may be replaced when the task allows it.
Do not inspect parent directories, benchmark harness files, grader code, reference images, or any oracle material.
Do not call write endpoints belonging to a reference site. Keep forms, counters, and similar behavior local.

Dependencies install offline from a vendored cache that matches the starter lockfile.
Do not add dependencies; the install will fail and the network is not yours to use.

`tools/screenshot.mjs` renders your work in the browser build the evaluator uses:

    node tools/screenshot.mjs --url http://127.0.0.1:4173/ --out shots/desktop.png \
      --width 1440 --height 900 --full-page

Before finishing:
1. Run the requested build command.
2. Start the app and capture every target viewport in the task with that tool, then read the captures back.
3. Exercise the required interactions and inspect the console errors and blocked requests it reports.
4. Write `submission/submission.json` using the contract in the implementer skill.

The harness will evaluate the filesystem after this process exits. Do not leave a dev server running.

Prompt rendered by the Harness for Claude Code.

/design-case-implementer

You are the candidate in ReconBench task `stackedit-field-notes` (Reconstruct a local-first Markdown workspace).

Read `task/brief.md` and every other file under `task/` before changing code.
Implement the result only under `submission/`. The starter may be replaced when the task allows it.
Do not inspect parent directories, benchmark harness files, grader code, reference images, or any oracle material.
Do not call write endpoints belonging to a reference site. Keep forms, counters, and similar behavior local.

Dependencies install offline from a vendored cache that matches the starter lockfile.
Do not add dependencies; the install will fail and the network is not yours to use.

`tools/screenshot.mjs` renders your work in the browser build the evaluator uses:

    node tools/screenshot.mjs --url http://127.0.0.1:4173/ --out shots/desktop.png \
      --width 1440 --height 900 --full-page

Before finishing:
1. Run the requested build command.
2. Start the app and capture every target viewport in the task with that tool, then read the captures back.
3. Exercise the required interactions and inspect the console errors and blocked requests it reports.
4. Write `submission/submission.json` using the contract in the implementer skill.

The harness will evaluate the filesystem after this process exits. Do not leave a dev server running.

Not publicly reproducible in this Pilot

Starter workspace: access restricted
Injected Skills: access restricted
Harness command: withheld until a runnable public bundle is released
Sealed evaluator: never Candidate-visible

Appendix ATask brief — exactly what the candidate agent receives

Field Notes Markdown workspace

Reconstruct the frozen local-first Markdown workspace shown in the supplied evidence. The target is a compact editing tool, not a marketing page. Match the desktop split editor/preview at 1440x900 and the phone editor layout at 390x844.

The initial document title is Field Notes and its exact Markdown fixture is:

# Field Notes

A calm place for **working drafts**.

## Today

Capture the layout.

Review responsive behavior.

| View | Mode | Status |
| --- | --- | --- |
| Desktop | Split | Ready |
| Phone | Editor | Ready |

Everything stays in this browser.

sync: ready

The desktop workspace keeps a dark formatting toolbar, equal editor and rendered preview panes, a narrow preview control rail, and a blue status bar. The phone layout hides the preview pane without shrinking the editor off-screen and keeps the reduced toolbar, title, preview controls, and status bar usable.

Implement these observable behaviors:

Reader mode changes the phone from editor mode to the rendered document shown in evidence/mobile-reader.png.
Editing Markdown updates the rendered preview immediately.
The current title and Markdown are stored locally and survive a same-origin reload. The persistence check replaces the title with Local Draft and the note with # Field Notes\n\nUpdated locally., reloads, verifies both saved values, then restores the title to Field Notes before comparing the state with evidence/desktop-persisted.png.

Expose the accessible name Markdown editor directly on the editable textarea or contenteditable element. Expose Rendered preview, Reader mode, and Document title on their corresponding controls and surfaces. Keep all behavior local. Do not embed StackEdit, copy its source, use a screenshot as the implementation, call production APIs, or load runtime assets from the network.

Run npm run build, then write submission/submission.json according to the supplied contract.