The Independent Game Studio

Turn the production brief into an in/out scope and review sequence. Preserve the playable reference and treat the synthetic records as a rehearsal, not validation.

Inputs: starter/case/production.json, starter/ops/release-checklist.md

Outputs: deliverables/first-shift-scope.md

Depends on: No prerequisites

• Names the two-day cap, all seven team owners, and the no-network/no-new-mechanics constraints.
• Lists publication and real participant recruitment as separate owner-approved effects.

Run the local feedback utility and inspect every record. Attach counts and a coding note distinguishing author-generated feedback from observed research.

Inputs: starter/data/playtests.csv, starter/tools/feedback_report.py, deliverables/first-shift-scope.md

Outputs: deliverables/feedback-baseline.json, deliverables/feedback-coding.md

Depends on: lock-first-shift

• Reports eight records, five completions, five bridge stalls, and three undo discoveries.
• Explains that the sample cannot establish market demand or an actual usability improvement.

Write an exact first-minute interaction specification using the current four-beat bridge rule. Include blocked movement, standing on a closing bridge, depot reload, and undo.

Inputs: starter/design/rules.md, starter/case/production.json, deliverables/feedback-baseline.json, deliverables/feedback-coding.md

Outputs: deliverables/onboarding-spec.md

Depends on: read-synthetic-feedback

• A player can determine whether the next bridge entry is legal from visible text.
• Each rule has an example action/state pair and does not change the three level layouts.

Revise the original courier SVG and palette into a compact, self-authored atlas or symbol sheet for the specified states. Avoid external fonts and borrowed game IP.

Inputs: starter/art/courier.svg, starter/art/palette.json, deliverables/onboarding-spec.md

Outputs: deliverables/courier-atlas.svg, deliverables/readability-notes.md

Depends on: revise-bridge-onboarding

• Contains distinct depot, courier, undelivered beacon, delivered beacon, open bridge, and closed bridge symbols.
• Every state remains distinguishable in monochrome and all SVG content is local.

Review the existing Web Audio cue sheet against the onboarding specification. Specify short tones, volume bounds, and a no-audio fallback rather than importing audio.

Inputs: starter/audio/cue-sheet.csv, starter/game/game.js, deliverables/onboarding-spec.md

Outputs: deliverables/audio-map.json

Depends on: revise-bridge-onboarding

• Includes blocked, loaded, delivered, and won events with durations no greater than 180 ms.
• Requires a user opt-in and keeps every critical cue available in text.

Produce a revised engine/UI source pair from the working reference, incorporating the reviewed specification, symbol work, and optional cues. Preserve offline file opening and deterministic rules.

Inputs: starter/game/index.html, starter/game/game.js, starter/tests/game.test.js, deliverables/onboarding-spec.md, deliverables/courier-atlas.svg, deliverables/readability-notes.md, deliverables/audio-map.json

Outputs: deliverables/game.js, deliverables/index.html, deliverables/engineering-checks.json

Depends on: prepare-readable-art, specify-local-cues

• The supplied engine tests still pass, including solver coverage for all three boards.
• The revised HTML references only local files and supports keyboard restart, undo, wait, and level selection.
• Records tests actually run separately from any unrun browser/device checks.

Use the regression table on the reference and revised build. Record state traces for bridge closure, illegal movement, delivery, winning, reset, and undo; do not mark manual browser checks passed without observation.

Inputs: starter/quality/acceptance.csv, starter/tests/game.test.js, deliverables/game.js, deliverables/index.html, deliverables/engineering-checks.json

Outputs: deliverables/regression-results.csv

Depends on: implement-onboarding

• Every acceptance row has an actual result or explicit not-run status and reproduction steps.
• No wall collision consumes a beat and each board can reach its win state.

Inspect the candidate for network references, storage, external assets, focus behavior, optional sound, and non-color state cues. Use the checklist to separate automated evidence from browser and assistive-technology limits.

Inputs: starter/ops/release-checklist.md, deliverables/index.html, deliverables/game.js, deliverables/courier-atlas.svg, deliverables/regression-results.csv

Outputs: deliverables/offline-accessibility-review.md

Depends on: verify-game-regressions

• Reports no network-capable gameplay code or remote assets.
• Explicitly records whether keyboard-only, sound-disabled, narrow-screen, and screen-reader checks were performed.

Write a small prospective test plan, not fabricated results. Ask participants to load, cross a bridge, deliver, undo, and restart; define consent and the minimum evidence needed to compare against the synthetic exercise.

Inputs: deliverables/feedback-coding.md, deliverables/onboarding-spec.md, deliverables/regression-results.csv, starter/case/production.json

Outputs: deliverables/prospective-playtest.csv

Depends on: verify-game-regressions, revise-bridge-onboarding

• Contains five observable tasks with neutral facilitator wording and completion/error fields.
• Identifies all new observations as pending and requires approval before recruitment or data collection.

Review the implementation, QA evidence, offline/accessibility review, and prospective test plan. Choose hold or a bounded candidate recommendation; do not publish or claim real player outcomes.

Inputs: deliverables/first-shift-scope.md, deliverables/regression-results.csv, deliverables/offline-accessibility-review.md, deliverables/prospective-playtest.csv, starter/ops/release-checklist.md

Outputs: deliverables/release-decision.md

Depends on: audit-offline-release, plan-observed-playtest

• Lists unresolved risks and every failed or unrun case.
• Any publication recommendation remains conditional on explicit owner approval and does not mislabel the seed as a shipped commercial game.