Flywheel v0.1.0

• fix(11-08): allowlist the 11-22 header reads and make the README build command work for a stranger
• Merge origin/main into the phase 11 branch
• chore(gsd): record use_worktrees=false, the setting that serialized phase 11 execution
• docs(11-11): complete the stuck-task exits and breach hold plan
• feat(11-11): Retry and Skip on a stuck task in the wave plan sheet (D-REV-41)
• feat(11-11): an observed scope breach holds the automatic merge (D-REV-42)
• test(11-11): an observed scope breach must not merge automatically, red (D-REV-42)
• feat(11-11): retry_task and skip_task, human-only (D-REV-41)
• test(11-11): a stuck task can be retried or skipped by a human, red (D-REV-41)
• docs(11-06): complete the update banner plan
• feat(11-06): mount the update offer once in the shell (D-187)
• feat(11-06): the update banner and its hook (D-187, D-188)
• feat(11-06): updateCheck and updateInstall at the adapter seam (D-187)
• docs(11-07): defer signing and notarization, the owner declined the Apple purchase (D-REV-60)
• docs(11-22): complete the two pulled-forward finish-review findings plan
• feat(11-22): the session header names its board item (four-1)
• test(11-22): the session header names its board item, red (four-1)
• feat(11-22): draw the +/- sign column in the diff gutter (three-3)
• test(11-22): a diff row says which side it is on, red (three-3)
• docs(11-05): record the keygen and secret-set deviations, and the green ci.yml re-run
• docs(11-05): complete the updater plan
• feat(11-05): updater commands, endpoint, pubkey and signing env (D-186, D-188)
• feat(11-05): add tauri-plugin-updater pinned at =2.12.0
• docs(11-22): plan the two finish-review findings the owner pulled forward (three-3, four-1)
• docs(11-18): complete contracts and brand rule plan
• docs(11-18): summary for the contracts and brand rule plan
• fix(11-18): brand-rule.test.ts typechecks without @types/node, using the workspace's node-shims
• docs(11-18): D-REV-01, 13, 28 and 30 closed; the eight 11-17 rows outside D-197..199 handed to D-196 filing
• feat(11-18): the brand rule widened to seven uses and enforced by a test that reads the shipped bytes (D-199)
• fix(11-18): the task row tail prints a value or nothing and nothing renders below 11px (D-198)
• docs(11-18): AGNT-07 says what the product does, a destructive action is refused and the user is told (D-197)
• docs(11-02): the repository is HighKnow-ai/flywheel and public, so 11-05 compiles the org endpoint
• docs(11-17): complete finish-review second pass plan
• docs(11-17): register entries D-REV-11, 12, 14-19 and 26 drained with the reviewer outcomes
• docs(11-17): finish-review record, nine reviewer runs with a disposition and evidence on every finding
• fix(11-17): permission rules table keeps granted authority readable and revokes once (D-REV-17)
• test(11-17): capture the permission-rules table the group-five reviewer could not see (D-REV-17)
• fix(11-17): the finish reviewers' four fixable findings, each with a WebKit assertion
• test(11-17): capture the launch sheet's Review feedback section (D-REV-19)
• test(11-17): capture the board mounted in the real app shell (D-REV-12)
• docs(11-16): update state and roadmap for plan 16
• docs(11-16): summary for the fan-out and real-binary plan
• docs(11-16): D-REV-34 and D-REV-35 wiring halves closed in the register, shim coverage gap logged
• test(11-16): run the real claude binary under the planning argv and assert no permission_request (D-REV-35)
• test(11-16): drive a whole fan-out through the real sidecar with only the model faked (D-REV-34)
• docs(11-10): record the owner's adopt-and-auto-merge decision as D-REV-59
• docs(11-10): complete D-REV-39 and D-REV-44 drain plan
• docs(11-10): summary, and D-REV-39 and D-REV-44 resolved in the register
• test(11-10): pin that no desktop command probes the case flag under the database lock
• fix(11-10): build the new test's git command in core, name the causes when the flap test fails
• refactor(11-10): remove may_merge and hold_for_limit, which nothing calls
• fix(11-10): no git config under the database lock; the launch reads the observed case flag
• fix(11-10): adopt a hand-run task instead of relaunching it; Unicode case fold
• test(11-10): a dispatcher launch is attributed to the system actor
• fix(11-10): a retried merge that fails the same way stays one record
• test(11-10): a persistently failing merge appends one event and does not flap
• docs(11-04): complete tag-driven release pipeline plan
• fix(11-04): cap the release notes at 200 commits
• fix(11-04): permission.test cleanup kills its own descendants, not every fake-claude
• fix(11-04): a zombie grandchild is terminated; a live one still fails the stop test
• fix(11-04): wait for the stubborn session's grandchild to leave the process table
• fix(11-04): scope-marker contract tests follow the spawn_prepared_run extraction
• feat(11-04): tag-triggered unsigned release workflow (D-182, D-185, D-186)
• feat(11-04): the git tag is the release version in all three manifests (D-184)
• docs(11-03): update state and roadmap for CI and open-core plan
• docs(11-03): complete CI clean-clone build and open-core proof plan
• docs(11-03): README building-from-source, sidecar tests, licence and architecture pointers
• feat(11-03): open-core proof, build and test with the relay seam deleted (DIST-04)
• feat(11-03): clean-clone build and bundle on every push (DIST-03)
• docs(11-15): complete fleet live reads, row density and jumps plan
• docs(11-15): route D-REV-47/48/49/50/51/54/55 in the register, log two session findings
• test(11-15): fleet row jumps with real key presses in WebKit (D-REV-51)
• test(11-15): assert where a fleet row lands, at the AppShell (D-REV-51)
• test(11-14): correct rate-limit provenance, a real hard limit is observed in transcripts
• test(11-15): measure the fleet surface and worktree inventory in WebKit, both schemes (D-REV-47, D-REV-54)
• test(11-15): one harness drives four fleet reads against a scratch database (D-REV-48/49/50/55)
• docs(11-14): update state and roadmap after plan 14
• docs(11-14): summarize three register items measured
• docs(11-14): route D-REV-32/33/36 in the register, log the Approve-error defect
• test(11-14): every dispatch state fits and is keyboard-reachable in WebKit (D-REV-36)
• test(11-14): measure the scope-denied feed row and refusal badge in WebKit (D-REV-33)
• test(11-14): freeze rate-limit wire shapes with recorded provenance (D-REV-32)
• docs(11-13): update state and roadmap after plan 13
• docs(11-13): summarize six deferred-review items measured
• docs(11-13): record what the six measurements closed and the residual human judgements
• feat(11-13): narrow the epic lens's headers to the epics the core's filter returns rows for (D-REV-27, A-04)
• test(11-13): the epic lens must narrow its headers under the has-worktree chip (D-REV-27)
• fix(11-13): give row-enter an explicit 'to' so the group wipe grows instead of stepping (D-REV-23)
• test(11-13): sample an expanding group's wipe at fixed times; it steps instead of growing (D-REV-23)
• test(11-13): a staged challenge row stays inside its 48px box in WebKit, also with an error beside it (D-REV-10)
• test(11-13): the uncommitted-changes Dot clears 3:1 on every row surface, ink read from the components (D-REV-09)
• fix(11-13): wrap a criterion row's actions below its statement in a narrow host (D-REV-07)
• test(11-13): the criteria rail at 900px must not clip its controls (D-REV-07)
• test(11-13): measure every token leaf at 4.5:1 with a diff row gutter-selected (D-REV-06)
• docs(11-12): update state and roadmap after plan 12
• docs(11-12): summarize forge ordering, conflict-exit focus and overlap toast
• feat(11-12): lift the overlap toast into a component and surface it from the merge sheet's launch (D-REV-43, D-REV-46)
• test(11-12): the overlap warning and a refusal reason must reach the screen from every launch site (D-REV-43, D-REV-46)
• fix(11-12): focus the first conflict exit when a merge result renders (D-REV-02)
• test(11-12): focus must land on a conflict exit when the merge result renders (D-REV-02)
• feat(11-12): extract the draft-hash-before-push ordering into core (D-REV-03)
• test(11-12): an epic request with a moved draft must never reach the forge (D-REV-03)
• docs(11-09): update state and roadmap after plan 09
• docs(11-09): summarize launch-path defects D-REV-37, 38, 45, 46
• fix(11-09): a launch that fails after its run row exists closes the run failed (D-REV-46)
• test(11-09): a launch that fails after its run row exists must leave no running row (D-REV-46)
• fix(11-09): a widen into a parked task records its edge by asking is_startable (D-REV-45)
• test(11-09): a widen into a task parked at pre-spawn attention must record its edge (D-REV-45)
• test(11-09): a live widen landing between decision and launch is refused (D-REV-37)
• docs(11-02): update state and roadmap after plan 02
• docs(11-02): summarize licence, architecture and owner decisions
• docs(11-02): add ARCHITECTURE.md with the Open-Core Module Boundary section
• docs(11-02): add Apache-2.0 LICENSE and NOTICE
• docs(11-02): record the owner's three one-way calls
• docs(11-01): update state and roadmap after plan 01
• docs(11-01): complete packaged sidecar tracer plan
• test(11-01): prove the packaged sidecar acks a ping from outside the source tree
• chore(11-01): untrack the generated sidecar launcher and declare the bundle contents
• feat(11-01): stage a bundled Node runtime, shim and launcher for the sidecar
• docs(11): create phase plan
• docs(11): create phase plan — 21 plans across 7 waves
• docs(state): record phase 11 context session
• docs(11): capture phase context
• docs(state): correct progress counters after Phase 10
• docs(project): evolve PROJECT.md after Phase 10
• docs(phase-10): complete phase execution
• docs(10): phase verification report
• docs(10): code review fix report
• fix(10): WR-04 scope the live-run re-check to the batch reclaim
• fix(10): WR-08 keep the notify loop's blocking calls off the tokio worker
• fix(10): WR-09 make a failed Dismiss visible and stop truncating the batch
• fix(10): WR-04 WR-05 re-check for a live run at removal and name a half-done removal
• fix(10): WR-03 re-measure reaches worktrees past the measure cap
• fix(10): WR-11 report a size write that updated no row as an error
• fix(10): WR-10 stop two timing tests failing on a loaded machine
• fix(10): WR-02 bound the post-exit pipe drain in output_with_timeout
• fix(10): WR-01 detect the forge outside the database lock in the forge tick
• fix(10): WR-06 migration 0021 restores the review backlog 0020 acknowledged
• fix(10): WR-07 keep tool input out of the OS notification banner
• docs(10): code review report for fleet supervision phase
• docs(10-09): update state, roadmap and requirements
• docs(10-09): summary and deferred review item
• test(10-09): a batch on its own connection is visible on the shared one
• feat(10-09): select, confirm the bytes, and see what happened to each row
• feat(10-09): reclaim worktrees in a batch that re-derives every row
• docs(10-08): complete worktree inventory plan
• feat(10-08): the worktree inventory view, with a reason on every row
• feat(10-08): derive worktree reclaimability and measure disk on demand
• docs(10-07): complete OS notifications plan
• docs(10-07): summary and deferred review items
• feat(10-07): one notification switch in Settings
• feat(10-07): notify at the two transition sites, and report the watched run
• feat(10-07): a pure coalescer and suppression rule, with the plugin on the core only
• docs(10-06): complete fleet row jumps plan
• docs(10-06): summary and deferred review item
• feat(10-06): reach a fleet row's board item in any project
• feat(10-06): open an agent's live output from its fleet row, in the right shape
• docs(10-05): complete blocked queue plan
• docs(10-05): summary and deferred review item
• feat(10-05): give the blocked queue a door on the rail and a second view on the fleet
• feat(10-05): answer blocked agents in place, one keystroke to one answer
• feat(10-05): read every agent blocked at a permission request, across all projects
• docs(10-04): complete spend and dismiss plan
• docs(10-04): summary and deferred review item
• feat(10-04): show what each finished agent spent and let the user clear it
• feat(10-04): one acknowledge statement, widened past abandoned
• docs(10-03): complete activity line on the fleet row plan
• feat(10-03): show what each live agent is doing on its fleet row
• feat(10-03): read the end of a session log without reading the log
• test(10-03): session log tail read boundary cases (RED)
• docs(10-02): complete three folded defects plan
• fix(10-02): a failing forge resolution is attempted once per tick, not once per worktree
• fix(10-02): narration and a tool call on one stream line both reach the feed
• fix(10-02): a child exiting at its deadline reports its real status
• docs(10-01): update state, roadmap and requirements; queue deferred visual review
• docs(10-01): complete fleet tracer plan summary
• feat(10-01): open-fleet global binding across keystroke, palette and native menu
• refactor(10-01): collapse the run-status vocabulary to one module
• feat(10-01): fleet tracer, every agent across every project in one view
• docs(10): create phase plan
• docs(10): create phase plan
• docs(10): research phase domain
• docs(state): record phase 10 context session
• docs(10): capture phase context
• docs(phase-09): close phase after gap closure (verification deferred per CLAUDE.md)
• docs(09): add code review report
• docs(09-13): update state and roadmap for the launch-body conflict plan
• docs(09-13): summarise the launch-body conflict check plan
• feat(09-13): show the launch overlap warning Rust produced
• test(09-13): add failing tests for the launch sheet's overlap warning
• feat(09-13): the shared launch body consults the one launch-conflict rule
• test(09-13): add failing tests for the launch body's conflict check
• docs(09-12): update state and roadmap for the launch-conflict rule plan
• docs(09-12): summarise the one launch-conflict rule and window plan
• test(09-12): the spawn closure consults the launch rule, closing the window test
• test(09-12): add failing window tests for the decide-to-launch gap
• feat(09-12): extract the one launch-conflict rule from readiness
• test(09-12): add failing tests for the one launch-conflict rule
• docs(09-11): update state and roadmap for the widen record plan
• docs(09-11): summarise the widen record and declare guard plan

Opening Flywheel for the first time (this build is not signed)

Flywheel is not signed with an Apple Developer ID and is not notarized, so macOS blocks the app the first time you open it.

If you downloaded the .dmg in a browser, macOS says that "Flywheel.app is damaged and can't be opened" and offers only Move to Trash and Cancel. The app is not damaged. There is no Open button, so right-click and Open does not help.

To open it:

1. Drag Flywheel.app from the .dmg into /Applications.

2. In Terminal, remove the quarantine mark that your browser put on the download:

   xattr -cr /Applications/Flywheel.app

3. Open Flywheel as usual.

Only do this for a build you downloaded from this repository's releases page. The command removes a macOS safety check for that one app.

This was observed on one Mac (macOS 26). Other macOS versions may word the dialog differently.