@torsday/omnifocus-mcp 1.2.0 → 1.2.1

package/CHANGELOG.md

@@ -5,6 +5,28 @@ All notable changes to `@torsday/omnifocus-mcp` will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). See [ADR-0011](./docs/adr/0011-versioning-and-stability.md) for the explicit definition of breaking vs additive changes in this project.
 
 
+## [1.2.1](https://github.com/torsday/omnifocus-mcp/compare/v1.2.0...v1.2.1) (2026-04-30)
+
+**Summary** — A focused reliability patch for OmniFocus 4.x compatibility and cross-transport ID interoperability. Seven bug fixes address real failure modes surfaced since v1.2.0: JXA scripts now correctly handle OF 4.x's quirky `class()` exceptions on tag parents and containing projects; `byId` misses are mapped to typed `NotFound` errors instead of leaking the raw `-1728` osascript code; and four write-path operations (`task_create`, `task_duplicate`, `task_reorder`, `project_move`) are routed through OmniJS to guarantee ID interoperability across all transport paths per ADR-0019. The `parentId` subtask filter also works correctly again after a `tasks()` vs `flattenedTasks()` regression. No breaking changes; all v1.2.0 call shapes are unchanged.
+
+### Fixed
+
+- **JXA tag parent `class()` guard — OF 4.x exception safety** — `tag_list` and `tag_get` now guard the `parent.class()` call and per-element `tag.id()` calls against the `Can't convert types` exception that OmniFocus 4.x throws on certain specifier types. Previously these would surface as opaque JXA errors; they now degrade gracefully and return the tag without parent info rather than crashing the response. ([bcaefb9](https://github.com/torsday/omnifocus-mcp/commit/bcaefb9de9bdef4ce92876bf580eb2a48927c45f))
+
+- **`byId` miss mapped to `NotFound` at the JXA boundary (closes [#674](https://github.com/torsday/omnifocus-mcp/issues/674))** — JXA's `flattenedTasks.byId()` returns a specifier with error code `-1728` ("Can't get object") when the ID doesn't exist, rather than `null`. This raw code was leaking through to callers. It is now intercepted at the transport boundary and converted to the typed `NotFoundError` the rest of the stack expects. ([ec53b88](https://github.com/torsday/omnifocus-mcp/commit/ec53b88e9733a3eaa24568da896d8a3da5ff377c))
+
+- **`containingProject().class()` exception preserves `projectId`** — In OF 4.x, calling `.class()` on a real project specifier throws rather than returning a class name. A prior guard was catching the exception but resetting `projectId` to `null` as a side effect. Fixed: the exception path now correctly leaves `projectId` set to the project's ID. ([b40a4a0](https://github.com/torsday/omnifocus-mcp/commit/b40a4a0d77c0165c4d2ae6305ecfddf1efb49e18))
+
+- **`parentId` subtask filter returns direct children only (closes [#695](https://github.com/torsday/omnifocus-mcp/issues/695))** — `task_list` with a `parentId` filter was calling `flattenedTasks()` on the parent, which recursively includes all descendants. Corrected to `tasks()` so only direct children are returned, matching the documented behavior. ([2796b30](https://github.com/torsday/omnifocus-mcp/commit/2796b30b9cf82090c67a4cbb0631f9e19712aaaf))
+
+- **`task_create` routed through OmniJS for ID interoperability (closes [#680](https://github.com/torsday/omnifocus-mcp/issues/680))** — `task_create` previously used JXA, which assigns a different ID namespace than OmniJS. The returned task ID could not be reliably round-tripped through OmniJS write operations. Routed through OmniJS per ADR-0019 to guarantee ID interoperability across all transport paths. ([0c9959a](https://github.com/torsday/omnifocus-mcp/commit/0c9959abae2bb0e4afb0bdeb76bd60310ee141de))
+
+- **`task_duplicate` routed through OmniJS for ID interoperability (closes [#692](https://github.com/torsday/omnifocus-mcp/issues/692))** — Same cross-transport ID issue as `task_create`. `task_duplicate` now runs via OmniJS and returns an ID that is valid for all subsequent operations regardless of transport. ([972ee8c](https://github.com/torsday/omnifocus-mcp/commit/972ee8c79154f523554e7e642a0262df1502dacb))
+
+- **`project_move` and `project_create` routed through OmniJS + JXA folder readback fixed (closes [#681](https://github.com/torsday/omnifocus-mcp/issues/681))** — Both operations now run through OmniJS, fixing cross-transport ID interoperability failures. A secondary bug — JXA folder status/move setters silently no-oping — is also fixed. ([a1bf707](https://github.com/torsday/omnifocus-mcp/commit/a1bf70729738f3b1ec3534f0e82fdea7f72f5818), [509a6bd](https://github.com/torsday/omnifocus-mcp/commit/509a6bd4d4bdca71ae3d9b889f42ec44db55d195))
+
+- **`task_reorder` validates parent before mutating (closes [#676](https://github.com/torsday/omnifocus-mcp/issues/676))** — `task_reorder` was applying the reorder even when the supplied `taskIds` belonged to different parent containers, producing silent data corruption. It now validates that all task IDs share the declared parent before any mutation. ([dc28308](https://github.com/torsday/omnifocus-mcp/commit/dc283086946d9418f7ea194136d5e91cb0e148a2))
+
 ## [1.2.0](https://github.com/torsday/omnifocus-mcp/compare/v1.1.0...v1.2.0) (2026-04-29)
 
 **Summary** — The headline additions are **outbound webhooks** (a full HTTPS + HMAC delivery subsystem that fires when OmniFocus state changes), **macOS Calendar integration** via a Swift EventKit bridge (with new `omnifocus://calendar` and `omnifocus://agenda` resources that merge calendar events with the OF Forecast view), **decision-journal** support (record user judgment on tasks/projects so agent-driven scans stop re-litigating the same anomaly), **natural-language perspective authoring** (a new MCP prompt + `perspective_create`/`update`/`delete`/`evaluate_dry_run` tools), **`task_defer_smart`** (intent-bearing defer-date grammar so agents stop landing tasks on weekends or 11 pm), and **mutation testing** wired in as a release-time hard gate (Stryker, calibrated baseline, fails publish on regression). Several existing surfaces were tightened — `task_extract_from_image` moved its post-parse validation rules into the Zod schema for cleaner error envelopes, several batch tools gained `.describe()` coverage on inner fields, and a handful of read-side responses now pair human-readable names with opaque IDs for the same reason v1.1.0 introduced the convention. Two new ADRs lock the architectures (ADR-0016 webhook delivery, ADR-0018 calendar bridge). No breaking changes; all v1.0.x / v1.1.x call shapes are unchanged.
@@ -186,6 +208,12 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ## [Unreleased]
 
+### Fixed
+
+- **`moveProject` routes through OmniJS + JXA folder-readback survives OF 4.x quirk (closes #681)** — fourth fix in the [ADR-0019](./docs/adr/0019-cross-transport-id-interoperability.md) series. The first slice of #681 routed `createProject` through OmniJS but left `moveProject` on JXA, where `target.move({ to: folder.projects.end })` fails with "Attempted to move data objects to a nil container" on OmniJS-created project specifiers. New `src/scripts/omnijs/project_move.js` uses `moveSections([proj], destination)` and resolves the destination via `flattenedFolders.filter(...)` (or `library` for the root). Routing flips `moveProject: "jxa"` → `"omnijs"`. Separately, the JXA folder-readback path on every project script (`project_get.js`, `project_get_many.js`, `project_list.js`, `project_create.js`, `project_update.js`) carried the same broken `f.class() !== "document"` guard that #673 already fixed for tasks: `f.class()` throws "Can't convert types" on a real Folder specifier in OmniFocus 4.x JXA, so the readback was silently returning `folderId: null` for every project in a folder. Replaced with the nested-try-catch pattern from #673 — treat the throw as "real folder", treat a successful return of `"document"` as the only skip path. The moveProject integration test now passes; project reads through JXA correctly surface `folderId` again.
+- **`duplicateTask` routes through OmniJS for cross-transport ID interoperability (closes #692)** — third sibling fix in the [ADR-0019](./docs/adr/0019-cross-transport-id-interoperability.md) series after [#680](https://github.com/torsday/omnifocus-mcp/issues/680) (createTask) and [#681](https://github.com/torsday/omnifocus-mcp/issues/681) (createProject). JXA's `task.duplicate()` and `container.make({...})` produce transient specifier IDs that downstream OmniJS reads can't resolve. OmniJS's `duplicateTasks([source], position)` and `new Task(name, position)` produce clones whose `id.primaryKey` is interoperable with both transports. New `src/scripts/omnijs/task_duplicate.js` mirrors the JXA props-copy surface (name, note, flagged, defer/due dates, estimatedMinutes, sequential, tags) and resets completion state on the clone (matching the JXA contract). Recursive clones use `duplicateTasks` and walk the resulting subtree to clear inherited `completed` flags; non-recursive clones build a single fresh task via `new Task(...)` — naturally produces an uncompleted childless result. Routing flips `duplicateTask: "jxa"` → `duplicateTask: "omnijs"`. Three of four duplicateTask integration tests now pass (was 1 of 4). The recursive case partially passes — `descendantCount` correct, but its downstream `listTasks({ parentId })` assertion still trips on a separate pre-existing JXA filter bug where parentId returns grandchildren too. Will file a follow-up for that.
+- **`createTask` routes through OmniJS for cross-transport ID interoperability (closes #680)** — sibling fix to [#681](https://github.com/torsday/omnifocus-mcp/issues/681). Per [ADR-0019](./docs/adr/0019-cross-transport-id-interoperability.md), JXA's `Task(props) + push()` returned a transient specifier ID that didn't match OmniFocus's persistent `id.primaryKey`, breaking subsequent OmniJS-routed downstream operations (`moveTask`, `reorderTask`, `duplicateTask`) which use the persistent key. New `src/scripts/omnijs/task_create.js` mirrors the JXA props-set surface (parent-task / project / inbox positions, note, flagged, defer/due dates, estimatedMinutes, tagIds, sequential, completedByChildren) and produces a task whose ID round-trips correctly across both transports. Routing-table flip: `createTask: "jxa"` → `createTask: "omnijs"`. Five of the seven named integration tests in #680 now pass: `createTask with projectId places the task in that project`, `moveTask into a project updates projectId`, and four `reorderTask` variants. The three `duplicateTask` failures and the `reorderTask validation when reference has different parent` failure trace to separate root causes (filed as follow-ups). Caller wrappers, OmniJsTransport contract, router exclusivity allowlist, and the routing-domain unit tests all updated to reflect the move; concurrent-test JXA-write fixtures now demonstrate via `updateTask` (still JXA-routed) since `createTask` is no longer the canonical example.
+
 ---
 
 ## [1.0.0] — 2026-04-25