Changelog

All notable changes to taskwarrior.nvim (formerly task.nvim) are documented here. Format follows Keep a Changelog; this project follows Semantic Versioning.

[Unreleased] — v1.5.0

Removed — Python backend

• bin/taskmd (the optional Python CLI) — removed. Taskwarrior.nvim has shipped a pure-Lua backend as the default since v1.1; the Python copy was kept as a fallback and as a standalone CLI for shell pipelines. Maintaining two near-identical implementations was a continuous source of drift bugs (e.g. the v1.4.1 missing-binary cascade where the Lua Layer-B check correctly short-circuited but the Python fallback ran task again and surfaced [Errno 2]). All edge-case parser tests from the pytest suite (unicode, CRLF, 100K-char descriptions, malformed lines, tag/UUID variants, multiple-token-same-key) were ported to tests/lua/spec/parse_spec.lua before deletion.
• backend = "python" config option — removed (silently ignored if set).
• taskmd_path config option — removed.
• tests/test_taskmd.py, tests/test_taskmd_extended.py — removed. CI loses the pytest job; coverage is now ~380+ Lua assertions plus the e2e harness driving real task.
• :checkhealth taskwarrior no longer reports Python availability or a taskmd CLI at … line — those checks are gone.
• README "CLI usage" section dropped; :help taskwarrior-cli removed.

If you were running taskmd from a shell pipeline, the equivalent today is nvim --headless -u <minimal_init> -c 'lua print(require("taskwarrior.taskmd").render({...}))' -c 'qa!'.

Fixed

• Healthcheck no longer reports Taskwarrior data at ~/.task when the task binary is missing (the fallback display was misleading; the block is now gated on vim.fn.executable("task") == 1).

v1.4.1 (post-launch polish)

Polish release driven by feedback from the v1.4.0 launch (issues #1, #2) and the distribution research that flagged the silent-broken-feedback default and the tofu-square first-touch UX. Every shipped change has a regression test under tests/lua/spec/; full Lua suite is now 219 assertions, all green.

Added — onboarding

• :TaskTutor — interactive 5-lesson tutorial covering the Taskwarrior CLI from scratch and graduating users into the plugin's buffer-as-database UX. Fully sandboxed: every task command runs against a throwaway DB in /tmp with rc.data.location= and rc.hooks=off flags; the optional [Tutor shell] terminal split spawns bash --norc --noprofile with TASKDATA/TASKRC env scoped to the temp dir. Structural isolation invariants enforced by tests/lua/spec/tutor_isolation_spec.lua (21 assertions).
• :TaskTutor reset — ends an active session and scans /tmp for orphan *_tw_tutor directories left behind by prior crashes.

Added — easy feedback flow

• :TaskFeedback unbricked — works on default install. Pre-fix the form refused to open because feedback_endpoint defaulted to false. The GitHub-issue and clipboard paths now always work; HTTP "Send" appears only when feedback_endpoint is configured.
• :TaskFeedback last-error — opens the form prefilled with the most-recent ERROR captured by the plugin's WARN+ ring buffer. Also bound to <leader>tF (default-on, opt-out via setup({ feedback = { feedback_key = false } })).
• First ERROR per session has a Tip suffix appended: "press <leader>tF or :TaskFeedback last-error to report this." Once per session — not spam.
• g? inside :Task buffers opens the feedback form prefilled with the active filter / sort / group + a sanitized snapshot of ~50 lines around your cursor.
• Privacy guarantees enforced by tests:
  • Description content scrambled (alphanumerics → a) while structural Taskwarrior tokens (project:, +tag, due:, priority:, <!-- uuid:... -->) are preserved verbatim. Length preserved character-for-character so layout bugs reproduce.
  • Task counts bucketed (DP-style ranges like 101-500), never the raw integer.
  • Ring buffer captures only WARN/ERROR notifications routed through the plugin — other plugins' notifications are never observed.
  • $HOME paths scrubbed to ~/....
  • Nothing auto-sends; every action is an explicit choice in the post-:w prompt.

Changed — DEFAULT COMMAND PREFIX FLIPPED (BREAKING for v1.4.0 users)

• Default command_prefix is now "Tw" (was "Task" through v1.4.0). Every command renames: :Task → :Tw, :TaskAdd → :TwAdd, :TaskFilter → :TwFilter, :TaskTutor → :TwTutor, etc. (40 commands total.) Resolves the collision with Shatur/neovim-tasks which also registers :Task (issue #1).
• Want the old :Task* commands back? Override before plugin load:

  {
    "matthandzel/taskwarrior.nvim",
    init = function() vim.g.taskwarrior_command_prefix = "Task" end,
    config = function() require("taskwarrior").setup() end,
  }
  

• The :Tw namespace was verified safe before flipping: a public- plugin survey found no Neovim plugin owning bare :Tw or any of the 40 :Tw* names. The only adjacent plugin is folke/twilight.nvim which uses :Twilight* — different command names, just shares the :Tw<Tab> completion menu.
• Collision detection. setup() scans for an existing :<prefix> and emits one clear WARN that names the override mechanism. No silent override. Suggests 'TaskW' as an example override (avoiding 'Tw'/'Task' since those are the meaningful defaults).

Added — configurable command prefix (issue #1)

• vim.g.taskwarrior_command_prefix and setup({ command_prefix = "..." }) accept any string matching ^[A-Z][A-Za-z]*$. vim.g wins if both are set (it's the only path that affects the lazy entrypoint in plugin/taskwarrior.lua, which runs before setup()).
• Help text in :TwHelp, lesson bodies in :TwTutor, the buffer- header read-only WARN, the error-notify hint, the feedback form preamble — all now interpolate the configured prefix at render time. No user-visible string lies about which command name to invoke.

Fixed

• task binary missing crash (issue #2). Pre-fix, opening :TaskAdd on a system without Taskwarrior installed raised E475: Invalid value for argument cmd: 'task' is not executable from inside vim.schedule — surfaced as an unintelligible Lua trace. Now: clear WARN at startup if task isn't on PATH, and taskmd.run() short-circuits cleanly so subsequent calls return "", 127 without raising. Both paths covered by tests/lua/spec/degraded_env_spec.lua. CLAUDE.md gains a new "Degraded environment — missing or broken hard dependencies" test tier so any new external-binary dep added in the future ships with a parallel degraded-env spec.
• Tofu [ ] square in non-nerd-font terminals. icons = true default now means auto-detect via vim.g.have_nerd_font rather than force-NF. Without nerd-font, the literal - [ ] shows through (matching what the parser expects). Existing users who set vim.g.have_nerd_font = 1 are unaffected. New escape hatch icons = "force-nf" for users whose NF detection is broken.
• demo/render-all.sh syntax error in the size-check step. Invalid bash (for ... in ... 2>/dev/null) replaced with shopt -s nullglob. The render step itself was unaffected.

Old [Unreleased] section follows

Large feature push closing the gap against ribelo/taskwarrior.nvim, huantrinh1802/m_taskwarrior_d.nvim, and duckdm/neowarrior.nvim. See docs/feature-gap-analysis.md for the full competitor survey that drove this list, and docs/research/ui-polish.md for the UI redesign notes.

Fixed

• Save with no changes is a true no-op. Previously a clean :w on a taskmd buffer ran the apply pipeline anyway, forking task and emitting Applied: +0 added, ~0 modified, v0 done. Now zero-action saves short-circuit silently in both confirm and non-confirm modes; toast suppression also covers the :w! / force=true path. (#370)
• Checkbox NF glyphs were empty strings in icons.lua for checkbox_pending / checkbox_started / checkbox_done, so the fallback always took the ASCII branch. Codepoints populated.

UI polish — checkbox icons land

• Checkbox virt_text overlay. Each task line now paints a 5-cell overlay over the literal - [ ] / - [>] / - [x] showing 󰄱 / 󰐊 / 󰸞 (or whatever the user override resolves to). The markdown source is unchanged — the parser regex ^- \[([ >x])\] (.+)$ still matches, so round-trips through taskmd apply are bitwise identical. Disable with icons = false; configure individual slots via icons = { checkbox_pending = "...", checkbox_done = "..." }. Auto-skipped when vim.g.have_nerd_font is unset and no slot override is provided (the ASCII fallback would just redraw the source).

UI polish (task lines look 10x better)

• Nerd Font icon system. New lua/taskwarrior/icons.lua exposes a typed slot table (priority_h, status_started, due_today, urg_1..urg_8, etc.) with per-slot NF glyph + ASCII fallback. Auto-selects based on vim.g.have_nerd_font. Override via config.icons = { priority_h = "!!!" } for individual slots or icons = false to force ASCII.
• Sign-column priority + status glyphs. Active tasks get a play icon, priority levels get chevron glyphs, overdue tasks get an alert icon — all in the sign column, which doesn't shift any text columns and is automatically revealed on task buffers.
• Contextual relative-date virt-text. due:2026-04-21 gets a right-aligned chip showing today, tomorrow, in 3d · Fri, next Mon, in 2w, 3d overdue, etc. The literal date in the buffer is never rewritten — :w still round-trips. Refresh is automatic on buffer render (and will be on CursorHold in a follow-up).
• OVERDUE badge pill for any task past its due date — contrasting fg/bg so it dominates the margin. Off by default (overdue_badge = false); the relative-date label already says "Nd overdue". Enable for a high-contrast alarm effect.
• Urgency bar glyph. 12.3 becomes ▇ 12.3 — 8-band unicode block glyph whose color matches the urgency_colors breakpoints. Off by default (show_urgency = false); enable for the score. urgency_bar = true controls whether to prefix the number with the band glyph when urgency IS shown.
• Started-elapsed chip. Active tasks show 󰐊 14m, 󰐊 1h32m, etc. in the right-align virt-text. (DST-safe: os.time() with isdst=false to match os.date("!*t") convention.)
• Palette refactor. Six semantic roles (TaskAccent, TaskUrgent, TaskWarn, TaskInfo, TaskTagHL, TaskSubtle). Field groups (TaskPriorityH, TaskProject, …) now link= to a role so a colorscheme override of the role recolors every field that uses it.
• TaskPriorityL is no longer green. Green reads "good / ready", but L priority is "ignorable" — now linked to TaskSubtle.
• TaskCompleted renders with strikethrough, making finished tasks unambiguous at a glance.
• TaskProject is now subtle, not bright teal — projects are always present, so they belong in the background.
• New config knobs: icons, urgency_bar, relative_dates, relative_date_refresh_ms.

Added

Task-level operations

• :TaskAppend / :TaskPrepend — add text to the description of the task under the cursor without entering the full modify flow. Buffer keymaps >> and <<.
• :TaskDuplicate — copy the cursor task as a fresh pending task (same project/tags/due, new UUID). Buffer keymap yt.
• :TaskPurge [filter] — irreversibly drop deleted tasks from the Taskwarrior database. Confirms before acting. Buffer keymap dD.
• :TaskDenotate — remove an annotation from the task under the cursor (counterpart to ga which adds one). Buffer keymap gA.
• :TaskModifyField <name> — field-specific pickers for project, priority, due, and tag. Reuses existing values from pending tasks so typos in project names don't create orphans. Buffer keymaps MM (project), Mp (priority), MD (due), Mt (tag).
• :TaskBulkModify <spec> — apply the same modify spec to every task in a visual/line range, one subprocess per task.
• :TaskLinkChildren / :TaskUnlinkChildren — mark - [ ] lines indented directly under the cursor task as depends: of it (or remove those dependencies).

Reports & workflows

• :TaskReport <name> — open a named Taskwarrior report. Built-in names mirror task <report>: next, active, overdue, recurring, waiting, unblocked, ready, blocked, completed, today, week, noproject.
• :TaskInbox — GTD-style triage of tasks added in the last 24h with no project, no due date, and no tags. Walks them one at a time with: set project / schedule / tag / defer / drop / skip / quit.

Live query blocks in markdown

• <!-- taskmd query: FILTER | sort:X | group:Y --> / <!-- taskmd endquery --> — any markdown buffer can host one or more live Taskwarrior views. Blocks auto-refresh on BufReadPost / BufWritePost (and on demand via :TaskQueryRefresh). Edits inside the block are not written back — blocks are read-only mirrors.

Visualisation & UI

• :TaskFloat [filter] — open a task buffer in a centered floating window. q dismisses.
• gf is now a bordered float, not a split (less disruptive).
• :TaskGraph — render the depends: graph as a Mermaid flowchart in a markdown code fence (renders in markdown-preview.nvim, quarto, and most obsidian-style viewers).
• Header stats virtual-text slots. header_stats = { fn1, fn2, … } renders right-aligned strings on the task-buffer header (each fn receives the full task list and returns a string or nil).
• :TaskExport [path] — write the current task buffer out as clean markdown (UUIDs and header comment stripped).
• :TaskSync — async task sync wrapper with progress, error detection (no-server / auth failure hints), and retry prompt.

Dashboard widget

• require("taskwarrior.dashboard").top_urgent(n) — returns a list of pretty-printed urgent tasks for alpha.nvim / dashboard.nvim startup sections. See |taskwarrior-dashboard|.

Telescope picker actions

• <C-d> delete with confirm, <C-y> yank UUID, <C-a> open quick-capture, <C-c> run arbitrary task <uuid> <verb>. Added to lua/telescope/_extensions/task.lua.

Configuration knobs

• tag_colors — per-tag highlight overrides (string or inline nvim_set_hl table spec). See |taskwarrior-tag-colors|.
• urgency_colors — user-configurable urgency→highlight breakpoints. Replaces the hardcoded 8/4/0 bands in virtual text and views. See |taskwarrior-urgency-colors|.
• notifications — per-category vim.notify gate (start, stop, modify, apply, review, capture, delegate, view, error, warn). Every notification in the plugin now goes through a single helper (lua/taskwarrior/notify.lua) that honors this table. See |taskwarrior-notifications|.
• granulation — opt-in auto-stop of running Taskwarrior timers after N ms of nvim-wide idle. See |taskwarrior-granulation|.
• header_stats — list of stat-slot functions rendered as virtual text on the task-buffer header.
• projects extended form — each project entry can now be a table with { name, view, filter, sort } instead of a plain name. The named saved view is auto-loaded when :Task opens from that cwd.

Changed

• gf (show task info) renders in a bordered float with q / <Esc> to close, rather than a split window.
• Virtual-text urgency numbers are now colored by urgency_colors breakpoints (previously always Comment).
• views.lua respects urgency_colors when rendering task lines.

Fixed

• +ACTIVE / +OVERDUE / +BLOCKED / +READY / +WAITING (virtual tag) filters silently returned empty in the Lua backend. The tag normalizer was rewriting +ACTIVE → tags.has:ACTIVE, but virtual tags aren't stored on tasks — only on-the-fly computed — so the tags.has: filter never matched. lua/taskwarrior/taskmd.lua now passes the full Taskwarrior virtual-tag set through verbatim. This repairs :TaskReport active/overdue/ready/waiting/unblocked/blocked, :TaskFilter +ACTIVE, and anyone else who filtered by a virtual tag.
• :TaskGraph produced Mermaid output that several extractors rejected. Node IDs are now t_<uuid8> (guaranteed letter prefix); all nodes are declared before any edges; labels sanitize every Mermaid-reserved character (| ; # { } [ ] " \``) before interpolation; empty DBs render a (no tasks)placeholder rather than an invalidflowchart TD` with no body.
• :TaskExport crashed with bad argument #2 to 'insert'. string.gsub returns (result, count) in Lua; passing it directly to table.insert tricks it into the 3-arg form. Wrap to coerce to string-only.
• :TaskUnlinkChildren only removed the first child. Taskwarrior's depends:-a,b parser reads it as "remove a, ADD b" — we now prefix every UUID with - so the whole list is removed.
• housing+food painted +food as a tag. syntax/taskmd.vim matched +\w[-_\w]* with no word-boundary check before the +. Fixed with \%(^\|[^0-9A-Za-z_]\)\zs lookbehind. (The Lua extmark-based highlighter in buffer.lua already handled this; the vim syntax file was a second, out-of-sync highlight layer.)
• priority:H was painted as generic taskmdField not taskmdPriorityH. Vim syntax rule-precedence is "later-defined wins at equal start position"; the generic field:value rule was defined before the priority-specific ones and stole the match. Rule order reversed.
• External Taskwarrior changes were silently clobbered on save. When Taskwarrior was mutated outside the plugin between render and save (CLI task add, mobile sync, another editor), the old save path would: mark the external add as done/delete, overwrite the external field change, or resurrect a task that was completed externally. The conflict detector existed (checked modified > rendered_at) but its output was ignored by apply.on_write. Rewrote compute_diff to implement a 3-way merge with five rules (external_modify / external_delete / external_add, plus the original modify/done cases). Added force propagation from :w! / config.force, and a structured conflicts list that surfaces to the confirm prompt (Apply safe / Apply force / Cancel) or aborts non-confirm saves until the user re-renders or forces. Python CLI mirrored for parity. Regression shield: tests/lua/spec/diff_external_changes_spec.lua (9 pure unit tests), tests/e2e/spec/external_changes_spec.lua (7 round-trip tests against a real task CLI), TestIntegrationConflicts in tests/test_taskmd_extended.py (5 Python integration tests).
• Right-aligned virt-text overwrote long task descriptions. right_align draws at the window's right edge unconditionally, so on wrapped lines it stomped on the last ~35 chars of the first wrap segment (e.g. project:career became proje7d overdue !OVERDUE). All right-align chips switched to virt_text_pos = "eol"; the chips now follow the content and never overwrite it. eol can push to an extra wrap line on long tasks — acceptable tradeoff.

Tests

• Added tests/lua/spec/features_spec.lua — 30 assertions covering module loading, config defaults and validation, query-block parsing, urgency_hl banding, per-cwd project entries, and report registry.
• Screen-rendering harness (geometric_overlaps in tests/e2e/spec/e2e_spec.lua): given a buffer, reads every right-align extmark and compares its width against the literal content width at a fixed window size. Flags any case where virt_text would overwrite visible characters on a real terminal. Catches the class of bug where "the test passed but the user saw garbled text" because extmark-data checks alone can't see layout conflicts.
• Added tests/e2e/ — full end-to-end harness. Spawns a temp TASKDATA, seeds fixtures, then drives each feature headlessly and asserts observable effects: task export for mutations, mmdc for Mermaid output, nvim_win_get_config for floats, extmark inspection for highlight/virt_text, vim.fn.synID() / synIDattr() for the vim syntax layer, buffer :w round-trips for apply/undo.
• Full suite: 151 Lua unit + 60 Lua e2e + 358 Python = 569 tests, up from 474.

[1.3.0] - 2026-04-19

Renamed the plugin from task.nvim to taskwarrior.nvim and tightened a few user-facing surfaces in the process. Functional behaviour is identical to v1.2.0; this release exists to reduce ambiguity when users discover the plugin via the taskwarrior keyword.

Changed

• Repository rename. MattHandzel/task.nvim → MattHandzel/taskwarrior.nvim. Update your plugin spec.
• Lua module path. require("task") → require("taskwarrior") (and the same for every submodule: taskwarrior.config, taskwarrior.taskmd, …).
• Vim doc tag. :help task.nvim → :help taskwarrior.nvim. Section tags renamed from task-* to taskwarrior-* (taskwarrior-config, taskwarrior-views, …).
• Health check. :checkhealth task → :checkhealth taskwarrior. The Python check is now a warn (informational), not an error — the default backend is pure Lua, so Python is genuinely optional.
• Plugin data dir. stdpath("data")/task.nvim/ → …/taskwarrior.nvim/. Saved views and apply backups migrate automatically on first use.
• Projects file. stdpath("data")/task_nvim_projects.json → …/taskwarrior_nvim_projects.json. Migrates automatically.
• User autocmd events / namespaces / augroups renamed from TaskNvim* / task_nvim_* to Taskwarrior* / taskwarrior_*.

Unchanged (intentionally)

• :Task* user commands and the taskmd filetype.
• bin/taskmd CLI (still stdlib-only Python, still named taskmd).
• The Telescope extension is still registered as task (so :Telescope task tasks keeps working — the extension name is independent of the module path).

Migration

• One-line lazy.nvim spec update: change "matthandzel/task.nvim" to "matthandzel/taskwarrior.nvim" and replace require("task") calls in the config = function() … end block.
• All persisted state is migrated transparently.
• If you depended on the old TaskNvimRefresh User autocmd pattern (or the task_nvim_hl / task_views_hl namespaces) for custom integrations, update to the new Taskwarrior* / taskwarrior_* names.

Deprecation shim

require("task") and require("task.*") keep working during the transition — lua/task/ is now a shim directory that forwards to lua/taskwarrior/. A one-time deprecation notice is emitted the first time require("task") runs. Slated for removal in v1.5; please update your configs by then.

[1.2.0] - 2026-04-17

Big release: splits the 2300-line init.lua monolith into focused modules, adds the first real Lua test suite, introduces data-safety defaults, and fixes a handful of user-reported bugs.

Released as task.nvim v1.2.0. Paths below reflect the layout at that tag; in v1.3.0 the plugin was renamed to taskwarrior.nvim and the lua directory moved to lua/taskwarrior/.

Added

• Modular architecture. lua/task/init.lua is now 273 lines (down from 2264). Domain logic moved into buffer.lua, apply.lua, capture.lua, delegate.lua, review.lua, saved_views.lua, projects.lua, completion.lua, commands.lua, help.lua, validate.lua.
• Lua test suite. 121 assertions across 4 specs under tests/lua/ (parser, render, diff, config) using plenary.nvim's busted runner. Bootstrap via ./tests/lua/bootstrap.sh; CI runs it on every push.
• Validated setup. require("task").setup({...}) now rejects unknown keys with typo-aware suggestions, and type-checks every known key including nested delegate.*, urgency_coefficients, urgency_value_mappers, and filters / projects.
• Auto-backup of Taskwarrior data before every apply. Default auto_backup = true copies ~/.task to stdpath("data")/task.nvim/backups/<timestamp>/; rolling retention of the ten most recent backups.
• Distribution surface. plugin/task.lua entrypoint (:Task exists without explicit setup()), ftplugin/taskmd.lua, syntax/taskmd.vim, doc/task.txt (:help task.nvim).
• :TaskFeedback command: structured feedback buffer that posts JSON to a configurable endpoint or opens a prefilled GitHub issue.
• Community health. CONTRIBUTING.md, CHANGELOG.md, SECURITY.md, .github/ISSUE_TEMPLATE/, PULL_REQUEST_TEMPLATE.md.
• Lint configuration. stylua.toml, pyproject.toml (ruff), .editorconfig. CI lint job runs advisory checks.
• CI matrix. Ubuntu + macOS × Python 3.8 / 3.10 / 3.12; added help-tag smoke verifying doc/task.txt (now doc/taskwarrior.txt).

Changed

• delegate.flags default is now "" (was --dangerously-skip-permissions). The old default silently disabled Claude Code's tool-permission prompts for every user; opting in is now explicit.
• Views render with consistent task-line coloring across tree, calendar, summary — shared render_task_line() helper.
• Urgency coefficients are now applied multiplicatively inside the Lua backend (via urgency_value_mappers) rather than being passed as rc.urgency.uda.FIELD.coefficient overrides to the Python CLI.

Fixed

• :TaskAdd no longer raises E565: Not allowed to change text or change window when nvim-cmp is installed. Close and submit are both deferred via vim.schedule so they don't run inside cmp's textlocked keymap solver.
• "Invalid buffer id" errors after :bwipeout on a task buffer. refresh_buf guards against stale bufnrs at entry, and the User TaskNvimRefresh autocmd body is wrapped in pcall.
• Triple backticks in a task description no longer paint every following task line as code. syntax/taskmd.vim clears markdown's multi-line code-block regions after inheriting.
• Smart j/k: screen-line movement that falls back to buffer-line when the cursor is blocked by a concealed UUID comment; window wrap = true prevents horizontal-scroll disorientation.

[1.1.0] - 2026-04-13

Added

• Pure-Lua backend (lua/task/taskmd.lua, default). Python is now optional.
• :TaskBurndown, :TaskTree, :TaskSummary, :TaskCalendar, :TaskTags — five read-only visualisation views.
• :TaskReview — guided urgency walk.
• :TaskDiffPreview — live virtual-text diff annotations.
• :TaskDelegate — hand a task (or a visual range of tasks) to Claude.
• :TaskSave / :TaskLoad — named views persisted to stdpath("data").
• :TaskFeedback — opt-in structured feedback buffer.
• Project auto-filter: :TaskProjectAdd, :TaskProjectRemove, :TaskProjectList.
• urgency_coefficients, urgency_value_mappers, and custom_urgency config knobs for UDA-aware sort.
• Nerd-font icons, configurable border style, open/transition animations, day-start-hour config.

Fixed

• Header-protection cache moved from closure-local to buffer-local, so :TaskFilter / :TaskSort / :TaskGroup followed by any edit no longer reverts the header.
• Buffer swapfile=false — no more stale .swp warnings on reopen.
• CLI refuses to apply a file with a missing/malformed header unless --force is passed — prevents the "every pending task marked done" failure mode when a user hand-writes a markdown file.

[1.0.0] - 2026-03-22

Released as task.nvim v1.0.0.

Added

• Initial public release. :Task, :TaskFilter, :TaskSort, :TaskGroup, :TaskRefresh, :TaskAdd, :TaskUndo, :TaskHelp.
• bin/taskmd CLI (Python, stdlib only).
• :checkhealth task (renamed to :checkhealth taskwarrior in v1.3.0).
• Demo GIF, README, MIT license.