Preserved from the deleted docs/roadmap.md. This records what tasks deliberately does not do and what was consciously postponed. Decisions that were made live as ADRs in docs/adr/; the command surface lives in README.md; the domain language lives in CONTEXT.md.
Out of scope (v1)
Explicitly excluded and not planned for the immediately following milestones:
- Multi-user collaboration, server-side sync, or any concept of remotes beyond
git push/pull if a human chooses to use them.
- Subtasks. The model is flat: Tasks with Dependencies, nothing else (see ADR-0004).
- Configurable column sets per project. The six Columns are fixed in v1 (see ADR-0003, ADR-0010).
- A TUI.
tasks board is a static print.
- Auto-dispatching Tasks to Claude Code sub-agents.
attendance signals eligibility; the orchestration loop is a separate concern.
- An agent-driven
tasks breakdown command that splits a Task into linked children.
- Enforced routing semantics for
attendance: unattended. In v1 it is a plain enum with no automated dispatch (see ADR-0008).
- Cross-project dependencies. Each Project is isolated.
- Time tracking, due dates, priorities, recurring tasks, urgency scoring.
- Encryption at rest.
Deferred (revisit before v2)
Considered during design and consciously deferred. Each has a v1 fallback and a trigger for revisiting.
- npm distribution. v1 ships GitHub releases with prebuilt macOS/Linux binaries (see ADR-0014). Revisit if users ask for
bunx tasks or platform-coverage gaps emerge.
- Auto-JSON output on non-TTY stdout and a
TASKS_JSON env var. v1 requires explicit --json. Revisit if agent ergonomics suffer.
- Agent-vs-human attribution (e.g.
TASKS_ACTOR, Co-Authored-By trailers). v1 inherits the user's git identity for all commits. Revisit when there's a real need to audit agent activity distinctly.
- Adaptive board layout (auto-stack on narrow terminals). v1 is fixed-width side-by-side; pipe through
less -S. Revisit if narrow-terminal use becomes common.
- Rich markdown rendering in
show. v1 uses light ANSI styling only. Revisit if Task bodies grow large enough that prose readability matters.
.tasks marker file for explicit Project-root pinning. v1 walks up to .git only (see ADR-0012). Revisit if users hit the no-.git fallback awkwardly.- Per-project config file (custom
done cutoff, custom column set). v1 uses fixed defaults and per-invocation flags. Revisit if projects need durably different settings.
- Hard transition gating (rejecting
mv ready when deps incomplete). v1 leaves transitions open; the Validator only protects the DAG (see ADR-0005). Revisit if open transitions cause observable problems.
- Whether
tasks next should hint at blocked Tasks instead of hard-filtering. v1 hard-filters. Revisit if agents would benefit from visibility into near-ready work.
- Configurable
done cutoff window. v1 fixes the default at 7 days with --all and --since <duration> as overrides. Revisit alongside the per-project config-file question.
Preserved from the deleted
docs/roadmap.md. This records whattasksdeliberately does not do and what was consciously postponed. Decisions that were made live as ADRs indocs/adr/; the command surface lives inREADME.md; the domain language lives inCONTEXT.md.Out of scope (v1)
Explicitly excluded and not planned for the immediately following milestones:
git push/pullif a human chooses to use them.tasks boardis a static print.attendancesignals eligibility; the orchestration loop is a separate concern.tasks breakdowncommand that splits a Task into linked children.attendance: unattended. In v1 it is a plain enum with no automated dispatch (see ADR-0008).Deferred (revisit before v2)
Considered during design and consciously deferred. Each has a v1 fallback and a trigger for revisiting.
bunx tasksor platform-coverage gaps emerge.TASKS_JSONenv var. v1 requires explicit--json. Revisit if agent ergonomics suffer.TASKS_ACTOR,Co-Authored-Bytrailers). v1 inherits the user's git identity for all commits. Revisit when there's a real need to audit agent activity distinctly.less -S. Revisit if narrow-terminal use becomes common.show. v1 uses light ANSI styling only. Revisit if Task bodies grow large enough that prose readability matters..tasksmarker file for explicit Project-root pinning. v1 walks up to.gitonly (see ADR-0012). Revisit if users hit the no-.gitfallback awkwardly.donecutoff, custom column set). v1 uses fixed defaults and per-invocation flags. Revisit if projects need durably different settings.mv readywhen deps incomplete). v1 leaves transitions open; the Validator only protects the DAG (see ADR-0005). Revisit if open transitions cause observable problems.tasks nextshould hint at blocked Tasks instead of hard-filtering. v1 hard-filters. Revisit if agents would benefit from visibility into near-ready work.donecutoff window. v1 fixes the default at 7 days with--alland--since <duration>as overrides. Revisit alongside the per-project config-file question.