The submission deadline is July 21, 2026 at 5:00 PM Pacific Time. The internal target is July 20 so the final day remains a submission buffer.
Deliverables:
- Product name and public GitHub repository.
- Durable engineering, privacy, and verification rules.
- Architecture decision and dependency audit.
- Reproducible development setup.
- Initial CI checks.
Exit gate:
- A fresh checkout can install dependencies and run all foundation checks from documented commands.
Deliverables:
- Desktop application shell with clear permission states.
- Grouped full-display and application-window source gallery.
- Record, stop, cancel, and playback flow.
- Local recording metadata and cleanup.
- A test fixture recording path.
Exit gate:
- A packaged development build records and plays back a workflow on macOS, and cancellation leaves no orphaned recording.
Deliverables:
- Bounded local frame extraction pipeline.
- Microphone-only voice note capture with typed fallback.
- Live transcription through
gpt-4o-mini-transcribe. - Structured learned-workflow proposal and optional schedule grounded in the transcript and frames.
- Editable goal, natural-language correction path, and inferred actions before anything is saved.
Exit gate:
- The reference recording and transcript produce a schema-valid recipe in deterministic and live-model tests, and a real WAV reaches the production transcription provider.
Deliverables:
- In-app API-key status, save, replace, and clear controls.
- Operating-system-backed encryption with no plaintext read-back path.
- Main-process credential resolution with app-key precedence and a development environment fallback.
Exit gate:
- A real Electron test saves an app key, proves the persisted file excludes its plaintext, resolves app-key status, and removes it successfully.
Deliverables:
- Versioned workflow intermediate representation.
- Editable recipe review UI.
- Capability and approval declarations.
- Deterministic filesystem dry-run engine.
Exit gate:
- The reference recipe previews exact renames and moves without changing the fixture, including collision and unsupported-file cases.
Verified result:
- The editable recipe persists locally and previews exact top-level moves or copies from dynamic learned file groups. Version 1 recipes migrate safely. Automated filesystem tests cover unmatched files, collisions, path traversal, changed sources, and late destination collisions.
Deliverables:
- Explicit approval of a persisted change plan.
- Real move and copy execution with exclusive destination creation.
- Durable local activity log with per-file outcomes.
- Honest completed, partial, and failed result states.
Exit gate:
- The Electron journey moves disposable footage, visual assets, and project packages into inferred folders, leaves unmatched files unchanged, and verifies the persisted activity log.
Deliverables:
- Scheduler for saved workflows.
- Run history with manual and scheduled results.
- Collision-safe reruns for the reference workflow.
Exit gate:
- Scheduled runs transform disposable fixtures correctly, and reruns do not corrupt prior results.
Verified result:
- Hourly, daily, weekday, and weekly schedules persist locally and resume checking when TaskTape opens.
- Scheduling is reviewed and saved with the workflow; spoken timing prefills the controls.
- The Scheduled inbox shows cadence, next run, last result, pause or resume, and Run now.
- A packaged Electron test forces a schedule due, moves a new asset fixture, and shows separate manual and scheduled history entries.
- Fresh plans, exclusive destination creation, and collision skipping protect reruns from overwriting prior results.
Schedule removal, rollback, and history search remain post-hackathon enhancements.
Deliverables:
- OpenAI computer-use execution for tasks that need an application or website.
- Native macOS screenshot, mouse, keyboard, scroll, and application-activation adapter.
- Turn limits, validated actions, untrusted-screen instructions, and safety-check stopping.
- The same save, run, schedule, and history lifecycle as deterministic tasks.
Exit gate:
- Deterministic agent and native-adapter tests pass, the Electron journey proves the complete task lifecycle, and one bounded live desktop task succeeds without an external or destructive action.
Verified result:
- Deterministic and Electron gates pass. An opt-in live test activated a disposable local browser page, used OpenAI computer use to click and type the exact verification phrase, confirmed the DOM value, persisted a completed run, and reached Run history in 14.6 seconds.
Deliverables:
- Editable expected outcome learned from the recording and user explanation.
- Schema-bound visual evaluator for the final replay screenshot.
- Passed, failed, and inconclusive results with screenshot evidence.
- Regression-focused completion and history views.
- Disposable broken and fixed browser target for a repeatable demo.
Exit gate:
- The same live learned check runs against broken and fixed target states, correctly reports a regression and a pass, persists both evidence records, and displays both in Run history.
Verified result:
- The paired OpenAI-driven desktop test passed in 1 minute 42 seconds on July 18. The broken target was classified as failed because the saved asset became Uncategorized. The fixed target was classified as passed because the saved asset retained Video. Both final screenshots and verdicts were persisted and surfaced in history.
Deliverables:
- Loopback-only MCP server embedded in the TaskTape desktop process.
- Headed instrumented browser for local development URLs.
- Accessible browser action tools with trace, screenshot, console, and network capture.
- Deterministic compilation from a completed evidence session into a saved Replay check.
- Shared desktop state showing active agent sessions and agent-created checks.
- Copyable Claude Code and Codex connection commands.
Exit gate:
- A real external MCP client connects to packaged TaskTape, launches the disposable broken target, records the failing interaction, persists the evidence bundle, creates a visible check, and retrieves it through the protocol.
Verified result:
- The source and packaged-app journeys both passed. The packaged application exposed 12 tools, observed the Uncategorized regression, wrote a non-empty Playwright trace and final screenshot, compiled the session into a review-required check, and returned that check through
list_checks.
Current boundary:
- Agent capture supports local browser applications and one active session. Release builds include their browser runtime. General desktop instrumentation, remote staging URLs, CI workers, and direct GitHub issue ingestion remain post-hackathon work.
Deliverables:
- Complete onboarding and empty/error states.
- Automated and manual regression pass.
- Signed or clearly documented local build.
- Public README, architecture notes, demo script, and sub-three-minute video.
- Codex session feedback and submission materials.
Exit gate:
- A clean-machine rehearsal follows the README, the demo succeeds twice consecutively, and every submission requirement is present.
Verified result:
- The self-contained 0.2.0 DMG includes Chromium and the TaskTape product icon.
- A read-only mounted-DMG run completed the external MCP reproduction journey with fresh user data.
- The production landing page and stable GitHub download passed desktop, mobile, redirect, full-download, SHA-256, and disk-image checks.
- The release remains clearly labeled as an unsigned Apple Silicon beta until Apple Developer ID and notarization credentials are available.
Deliverables:
- Reddit competitor and practitioner research with explicit product boundaries.
- Evidence inventory on every agent-created check.
- Ticket-ready Markdown report copied without uploading local evidence.
- Ordinary Playwright TypeScript export built from observed structured commands.
- Updated desktop and mobile release site using a verified product screenshot.
Exit gate:
- Unit tests verify deterministic compilation and legacy-session refusal. The Electron MCP journey reads the report from the system clipboard and validates the exported test. That journey passes against the development build, unpacked production app, and read-only mounted DMG.
Verified result:
- The complete gate passed with 66 unit and integration tests, 10 Electron journeys, two release-site browser viewports, package integrity checks, and two packaged-app MCP runs. Exported tests use no model at run time and remain review-required. Automatic repair, direct issue-tracker OAuth, and managed cloud execution remain out of scope.
Only submission fixes, reliability fixes, and final upload verification. No new product scope.